Moonshot Weekly - Big Ideas for the AI Age.

The Big Idea

Pay to Play.

Nvidia agreed to acquire a French AI darling - Hugging Face, for $12.9bn this week (the biggest acquisition in its history), and only three years after Hugging Face was valued at $4.5bn.

Meanwhile last week, Stripe also agreed to acquire OpenRouter for $7.5bn, just months after it was valued at $1.3bn.

So what’s happening and why should we care?

It looks like big tech is now starting to make moves to acquire not only new layers of the tech stack, but also entire ecosystems around them, and there is a good reason behind the trend - read on.

Nvidia already dominates the hardware underneath AI, and with Hugging Face, and its 3m public models, 1m datasets and 1.4m spaces, it will now effectively be able to have control of one of the most important distribution layers sitting above it: open source AI.

What makes the timing even more strategic? Nvidia’s quarterly financial report, also published this week, shows that the company hit $96.2bn in quarterly revenue, with $89bn coming from data centres (!).

And so, paying $13bn for a critical layer above that core business looks like no-brainer.

Now, OpenRouter, which currently routes more than 10tn tokens a day across 400+ models and from 80+ providers, is a strategic vertical integration layer for Stripe too. - Stripe already processes payments via its token-based billing and now, with OpenRouter, it will be able to decide which model actually receives those tokens.

Think about it, with hundreds of models increasingly competing on price, latency and performance, the router will also eventually become a powerful control tool for Stripe AND a lucrative income stream opportunity.

As you may have guessed, Stripe and Nvidia are not alone..

CoreWeave in March last year agreed to pay $1.03bn for Weights & Biases, expanding into model training, evaluation and monitoring. Cloudflare bought Replicate in November last year, bringing a catalogue of 50,000+ AI models directly into its platform. Early this month, Dynatrace agreed to buy Arize for $915m to expand into full-lifecycle AI observability.

The common denominator across all these acquisition opportunities? Yes, the adjacency of the product stack, but more interestingly, the model neutrality.

OpenRouter’s value comes partly from letting developers move between 400+ models rather than locking themselves into one. Hugging Face became the default home for open models precisely because it sits across the ecosystem. Replicate, similarly, lets developers access models from different providers. See the common thread?

As the market is becoming more fragmented, (81% of enterprises now use 3+ model families in testing or production - up from 65% a year ago), neutrality carries now a lot of value.

Think about it, the company sitting between those models does not need to predict which model will win—it gets the traffic whichever way usage moves! That makes neutrality an increasingly lucrative position to be in, and explains why the largest players now acquire it at scale.

And that’s exactly why what’s really happening is not only a product-layer expansion play, but rather an ecosystem-level expansion play.

If Nvidia owns one of the main places developers discover and deploy open models, it also gains visibility into what is gaining traction and a potential route for directing workloads towards Nvidia infrastructure.

Similarly, if Stripe owns the router processing trillions tokens a day, it sees not only where money moves through AI, but, increasingly, where model usage itself moves.

.. and over time, that visibility, that data, may be almost as valuable as the revenue. Here you have it: another toll booth ticketing opportunity.

The beautiful thing is, once you own the junctions of the AI stack and ecosystems that live within them, you don’t need to win every layer … you get paid whichever way the traffic goes.

GTM Edge

Not the Average.

There have always been two universes of founders: the average and the extraordinary.

It also often takes an extraordinary person to spot the extraordinary, but perhaps the new uncomfortable truth is that it has never been easier to be average.

In the AI era, when tooling and access are democratised, the output starts to converge. The same product pages, the same language, the same logic. In fact, average today looks better than it used to. Average now has a better landing page. Better copy. Average ships more. Average sounds more sophisticated.

The side effect of ChatGPT has been that it reinforced the worldview of the average, averaged out those in doubt, and made it even harder (and rarer) to stay extraordinary.

The price?

That might surprise you. The damaging effects of the global rise of ordinary are actually slow to surface. Why? Because averaging out happens on both sides: to those who are judged and those who are judging. And so, the match might still happen (an investment, a contract, a pilot) but, you’ve guessed it, the outcome will be ..ordinary.

Some people are blessed to be born lateral thinkers. If you’re one of them, I hope you’re at least modestly aware of it.

If you’re not sure and want to give yourself the best chance of disproportionate outcomes, my thoughts are: in order to stay ahead of the consensus, maybe counterintuitively, build with ordinary in mind. Keep the pulse of where the market is. What is already legible. Where the consensus lies. What everybody knows. What sounds familiar. What gets the most likes — keep note of that!

The hard part? Treat the process with forensic coldness and accuracy. Don’t engage. Don’t confuse faster execution with insight. Looking current with being right. Don’t get pulled into the emotional process. Keep asking: ‘What is everyone missing?’, ‘Where is the narrative wrong?’. Do it continuously and diversify your sources.

Stay quietly rebellious. Quietly, because without becoming someone else’s reference point (!).

As you go through this open-ended process, as you learn and build, and as you need to engage with the average, don’t drift toward ideas that are easy to explain. Instead, spot and focus on those important for you to build, based on the knowledge you have.

Average founders use AI to get to a presentable answer faster. Don’t use AI to replace original thinking. Use it to buy yourself more time and confidence for original thinking.

Data

European Funding Up. And Up.

It looks like VC investment in Europe is on a rapid rise, the latest data finds.

European startups raised $24B in Q2 2026, the highest quarter in four years (!) and up 33% quarter-over-quarter.

AI companies raised just over $10B - that’s 40% of the entire deal value and also the highest quarterly total on record.

However, when you look over the Atlantic to see how these numbers compare with North America, the excitement somehow fades - because the gap is enormous. In H1 this year, European startups raised $42B (up 50% YoY), meanwhile North American startups raised $392B during the same period - that’s over 9x Europe’s total (!).

Geographically speaking, the UK leads the pack with 43% of all European funding ($10.4B) - just $400M short of its all-time quarterly peak ($10.8B, 2021), followed by Germany ($3.2B), France ($2.4B) and Sweden ($2B).

When it comes to round sizes and that $24B total, 42 companies raised $100M or more, capturing 65% of all European VC funding of Q2. Four companies (Isomorphic Labs, Neura Robotics, Ineffable Intelligence, Stegra), raised rounds of at least $1B, accounting for 25% of the entire quarter.

Late-stage companies raised $12.1B (up 90% YoY). .. Even seed was unusually concentrated - $1B of the $3.2B raised by seed companies went to just one company - Ineffable Intelligence.

Community

Female Innovation Index.

Swedish diagnostics startup Readily Diagnostics, co-founded by Liza Löf, secures a €811k grant from the Gates Foundation; Hungarian EdTech startup BOOKR Kids, co-founded by Dorka Horváth, raises a €6.1M Series A round led by TCEE Fund IV; Belgian construction tech startup BIMproFIT, co-founded by Sandy Jorens, secures a €900K funding round from imec.istart; UK industrial software startup Intropy, co-founded by Franziska Kirschner, raises a $11M round led by Felix Capital.

Weekly funding coverage from Female Foundry. View the 2026 Edition of the Female Innovation Index.

Events next week

Tuesday, September 01, Stockholm ➯ Security in the AI era | Wednesday, September 02, London ➯ Unruly Play, ➯ Tech for Good Meetup, Paris ➯ LP/GP IPEM Global Rooftop Cocktail | Thursday, September 03, London ➯ Run & Raise: Female Founders & VCs, Berlin ➯ AI-Native Engineering - Meetup, Amsterdam ➯ Hackers & Founders Meetup | Saturday, September 05, London ➯ The Startup Stroll.

List your event

That’s all for this week. Have a good Sunday.

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Big ideas for the AI Age.

The Big Idea

Everything Gained.

Revolut founder Nik Storonsky announced a new $500m VC fund this week, twice the size of the $250m vehicle it raised only 15 months ago, to back growth AI companies across fintech, healthtech and deeptech. Last week, Goldman Sachs announced doubling down on its new alternative-investments platform allowing wealthy individuals and family offices to invest directly into individual private companies, skipping conventional VC or private equity funds.

This is a big deal, and Goldman has been building the plumbing for this for a while.

In January it bought Industry Ventures, a $7bn venture platform with a two-decade-long experience investing across over 800 venture and tech funds and hundreds of individual companies. Also just a month ago, Goldman and T. Rowe Price launched something also interesting: an interval fund giving private investors exposure to private equity, private credit, infrastructure, real estate and late-stage private companies .. starting with just a $2,500 ticket (!).

So why should we care?

Because the amount of capital required to finance the AI boom is beginning to stretch well beyond what the traditional VC model is able to provide.

To give you an idea, global VC funds raised just $122bn in 2025, down from $414bn at the 2022 peak. At the same time, global VC investment reached $560bn in just the first half of 2026, driven overwhelmingly by enormous AI rounds.

Traditional VC is not disappearing, but it’s starting to become only one piece of a much larger capital stack - and this is where things get interesting.

Goldman is effectively opening new source of capital: private wealth. Instead of raising from a VC whose own capital comes primarily from pensions, endowments and institutions, startups can now raise directly from wealthy investors, making assets that were historically available almost exclusively to large institutions available to a much broader investor base.

And private wealth is not the only capital source.

Just two weeks ago, Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create financing platforms designed to mobilise more than $500bn of third-party capital for AI infrastructure. The really interesting part is how Nvidia describes what is being financed: “compute itself is becoming an investable asset, capable of generating long-duration, usage-linked revenues”.

That moves AI financing into completely different territory.

We are already seeing GPUs financed more like aircraft, power plants and other productive assets than venture-backed tech.

CoreWeave closed an $8.5bn non-recourse financing in March, secured by AI computing infrastructure, and getting investment-grade rating. Two months later, it raised another $3.1bn through the first publicly syndicated HPC-backed financing vehicle, opening GPU-backed loans to an even wider group of credit investors.

In July, Nebius raised $775m facility backed by deployed GPUs and contracted cash flows.

That is a pretty radical change in the economics of building an AI company. Instead of financing every new GPU with expensive equity, those companies can now sign a long-term customer, borrow against the contract and the hardware, and finance the next cluster largely with debt!

Even the most cash-rich companies in the world, the hyperscalers, are getting creative too, bringing infrastructure investors directly into the financing of their AI buildout rather than funding everything from their balance sheets. - Meta last year formed a joint venture with Blue Owl to finance its Hyperion data-centre campus.

Then, there are secondaries.

If AI companies can raise billions privately and remain private for longer, early investors and employees cannot simply wait indefinitely for an IPO.

Global private-market secondary transactions reached a record $226bn in 2025, up 41% in a year. Within VC and growth specifically, GP-led secondary transactions exceeded $12bn, growing more than 40% year-on-year. -Goldman’s new platform includes a dedicated secondary team.

There is, of course, a catch.

New financial structures do not make risk disappear; they simply move it around. GPUs depreciate. Customer contracts fall through. Secondary markets are still illiquid.

But the new AI financing direction is difficult to now miss.

The first AI boom was largely funded by venture capital and the balance sheets of Big Tech. The next one is likely to pull capital from almost everywhere: founders, family offices, private wealth, insurers, credit funds..

And if that continues, the biggest change of the AI era may not simply be what gets funded, but also who gets to fund it.

GTM Edge

Pilots Backwards.

Last week I wrote about validation, about what to focus on to get your business off the ground and investors excited. And, funny enough, early this week, I spoke with a team who already thought they had commercial validation.

On the surface, their traction looked impressive. They were in the process of running five pilots, big logos. But when I asked how many pilots had actually converted into, or were due to convert into, a paid deployment, the answer was somewhat vague (which is a different way to say: zero).

I’ve looked into the state of AI adoption recently (see more on that below) and on a good side, enterprises are running more AI experiments than ever, on a not-so-good-side, 83% of them converted fewer than half of their AI pilots into production over the past 12 months, and more than a third converted fewer than one in four.

The mistake I see founders make is assuming the pilot will naturally become a contract if their product works. “Naturally” is the key word here. Because “naturally” hardly ever happens! If you are really serious about commercial validation, you should design the pilot, getting your prospective clients not only to try your MVP, but also to integrate it, trust it, budget for it, and keep using it when the novelty wears off.

You might be surprised, but pilots usually die for reasons that have nothing to do with your product. They die because, for example, a security team was not involved early enough. Because the product does not fit into the existing workflow (and the internal team is hesitant to change it). Because the ROI is vague. Because nobody owns the next step (or is responsible for it). They die because the champion likes it, but the budget holder doesn’t care. Because implementation feels bigger than the pain.

And so, if “naturally” doesn’t work, the only way for you to think about pilots is to think about them backwards.

Let me explain. Before the pilot even starts, you should already know (and establish it with the prospective client sponsor and a budget holder), what success looks like, who signs off, what metric has to move, what system you need to integrate with, what risk needs to be cleared, and what happens if the pilot works.

Do not start with: “Try this and tell us what you think.”

Start with, e.g: “If we reduce ticket resolution time by 30%, who needs to approve rollout?” Or: “If we save your team 15 hours a week, what budget does this come from?” Or: “If this works in one region, what is the path to three more?”

A pilot without a production path is just unpaid consulting with nice branding!

A good pilot should have a clear use case, a clear owner, a clear success metric, a clear timeline, and a clear commercial next step. If any of those are missing, you are probably not running a pilot, but an experiment for someone else’s learning.

And in AI, that gets expensive fast! (see: Week 205)

That’s why the best founders I meet these days are not trying to collect logos of pilots they are running. They are trying to manufacture conversion.

They narrow the use case. They define the metric. They involve the buyer early. They price the pilot. They make security and integration part of the plan (not a surprise at the end). They treat production as the ultimate goal from day one - and they work from there.

Remember, there is an abundance of curiosity from enterprises to try AI products right now, but production is scarce.

Make sure you push your way backwards, learn from the process, and stay on the right side of that equation.

Data

A Tale of Two Worlds.

It looks like AI spending is on a rise — but the gap between the big-buck spenders and everyone else is exploding.

Between January 2024 and July 2026, AI spend per employee among the top 1% of companies jumped from $384 to $7,516 a month. That’s a 19.6x increase in just two and a half years.

At the same time the top 10% enterprise spenders followed a similar trajectory, going from $72 to $1,017 per employee per month — a 14.1x increase.

But when looking at a median company, the picture looks very different.

Its AI spend rose from just $3.29 per employee per month in January 2024 to $11.79 by July 2026. That is still a 3.6x increase, but nowhere near the acceleration happening at the top of the market!

And that the size of the gap really matters.

In January 2024, the top 1% were already spending roughly 117x more per employee on AI than the median company. By July 2026, that gap, however had widened to around 640x!

Even within the top end of the market, the separation is widening. The top 1% now spend more than seven times as much per employee as companies in the top 10%.

And so, when arguably everyone is spending more, only a relatively small group of companies is scaling its AI investment at an entirely new universe of scale.

If you take a view that AI spend is strategic and directly translates into productivity and value creation, the real divide may soon be less about who is using AI and much more about how aggressively they are deploying it.

Community

Female Innovation Index.

Danish healthtech startup Aisel Health, co-founded by Augusta Klingsten Peytz, raises a €1.7M pre-seed round led by Caesar Ventures; Serbian-founded B2B media / martech startup Astute, co-founded by Vida Stanić, raises a $1.2M pre-seed round backed by Flyer One Ventures; UK maritime robotics startup Oshen, co-founded by Anahita Laverack, raises a $5M round led by Lunar Ventures; Icelandic energy software startup SnerpaPower, co-founded by Íris Baldursdóttir and Eyrún Linnet, raises a €3.4M seed round led by Encevo; UK proptech / energy AI startup Camion, co-founded by Lorna Walker and Paige Mullen, has been acquired by Deepki.

Weekly funding coverage from Female Foundry. View the 2026 Edition of the Female Innovation Index.

Events

Tuesday, August 25, Berlin ➯ Berlin. Vibe Coding Club: Demo Night | Wednesday, August 26, London ➯ Summer Sundown Social @ the Hub | Thursday, August 27, Denmark ➯ Danish Entrepreneurs x TechBBQ, Berlin ➯ Deep Tech Night: Quantum & Photonics | Friday, August 28, London ➯ Work in Fintech AI Summit 2026.

List your event

Have a restful Sunday. See next weekend!

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Big questions for the AI Age.

The Big Idea

Europe’s AI Factories.

Demis Hassabis, cofounder and CEO of Google DeepMind, stepped down from day-to-day leadership last week. Demis has led DeepMind since founding it in 2010, and while he is not leaving Google—the move has led to arguably lab’s biggest talent reshuffle in 16 years.

Meanwhile, also last week, the chief scientist at Google Jeff Dean also announced his departure after 27 years to build a new AI startup, Discovery Loop, taking some serious talent with him.

Jeff and Demis are not alone.

Yann LeCun, the former Chief Scientist at Meta, left Meta end of last year to build Paris-based AMI Labs, which went on to raise a $1.03bn seed round this March (one of the biggest European seeds ever recorded). And that’s just top of the iceberg.

Add dozens of less famous researchers making similar moves and it increasingly looks like we are witnessing a generational wave of AI startup creation out of the Europe’s labs.

The numbers are pretty extraordinary. There are now at least 112 former DeepMind employees who had founded a startup, or appeared to be launching one, in just the previous 18 months!

38 had already started companies and another 74 had moved into “stealth” roles in new startups. 28 are building in the UK, 3 in Spain, 2 in Switzerland, 2 in Germany, 1 in Austria and 1 in Poland. That is at least 37 new companies across Europe from one alumni network in just over a year thats is also fuelling Europe’s AI valuations!

To name a few, just this year, Ineffable Intelligence raised $1.1bn at a $5.1bn valuation even before having a commercial product in April, Recursive Superintelligence, raised $650m at a $4.65bn valuation also in April, finally, Mistral is rumoured to be raising €3B at €20B valuation later this year.

DeepMind has clearly become a top European AI startup factory.

And DeepMind is not the only one.

About 350 kilometres away in Paris, Meta has quietly also built another European AI factory. FAIR Paris opened in 2015 and has since hosted more than 120 researchers and PhD students. Of its alumni, 52% joined or founded European AI startups. Not surprisingly, two of Mistral’s three cofounders (Guillaume Lample and Timothée Lacroix) came from FAIR Paris, as did three of Kyutai’s six founders.

Then there is Palantir, which represents a slightly different kind of AI startup factory, producing not only frontier-model founders but also engineers that understand well how to deploy AI inside governments, defence and large enterprises… By early 2025, 19 former Palantir employees in Europe had built startups of their own, including ElevenLabs, Arondite, Bastion, Fern Labs and Wondercraft.

So what happens now that the biggest technology shift in decades collides with a continent that is finally accumulating these factories?

Mistral is only three years old. ElevenLabs is four. Recursive Superintelligence about one year old. Eventually, some of their best engineers will leave. So will people from Helsing, Synthesia and Legora. Some will become founders. Others will become angels and VCs. That is how tech ecosystems compound: one successful company produces ten companies, those ten eventually produce another hundred.

There is, however, one rather large catch.

DeepMind may be Europe’s most productive AI startup factory, but Europe is not necessarily keeping what it produces.

Of those 112 recent DeepMind alumni mentioned earlier, about 60% are building in the US - the biggest destination for European AI talent.

And the wider numbers point in the same direction. 6 of Europe’s 15 largest AI employers are US Big Tech companies (think: Google, Microsoft, Meta, Amazon, Apple and Nvidia). 149 European-origin unicorns have already relocated to North America, versus just three moving the other way.

And so, Europe increasingly has the talent and increasingly has the factories. What it still has to prove is whether it can keep enough of the companies those factories produce.

That makes the Hassabis transition particularly interesting.

The question for Europe right now is not simply whether more senior DeepMind researchers leave after him. Given the mind-blowing valuations and capital available to back elite AI researchers, more probably will.

The much more important question is - where they will build next?

GTM Edge

On Validation.

I spoke with a founder this week, who’s been working on an agent validation solution and preparing to fundraise.

When I asked what he wanted to validate with the first round, he pointed to almost everything: the model, the UX, the onboarding, the integrations, the ICP, the tech roadmap, the team.. all useful but (to my slight disappointment) he never mentioned commercial validation.

And that is the part I think too many AI founders are underestimating right now.

When you’re building an AI startup, it’s almost dangerously easy to build something impressive quickly but still having no business.

How do you know whether you have a business? Not just by knowing that your product works, but by knowing someone (ideally multiplied) needs it badly enough to pay for it now, from a real budget, and with real urgency.

That is commercial validation.

And if you are raising your first round, this should be one of the main things you are trying to prove!

The mistake I see founders make is to treat commercial validation as just a sales problem, when actually it is a company-wide problem! It tells you who the buyer is, what the wedge should be, how urgent the pain is, what budget you are entering, how to price it, what to build next, and whether the market is pulling or just politely engaging.

Remember, a prospect saying ‘this is cool’ is NOT validation. A user saying ‘we’d love to try this’ is NOT validation. and a free pilot is WEAK validation. Even usage can be misleading, if nobody can explain who would pay, why they would pay, and what budget it would come from.

See, if you really want to nail your commercial validation, you need to be far more specific.

Can you name the buyer, not just the user? Can you say what budget line this comes from? Can the customer explain what they would stop doing, stop paying for, or avoid hiring because of your product? Can you identify the workflow you replace? Can you point to the moment where the product creates enough value that payment becomes a natural next step?

That’s the difference between interest and intent.

One more thing: don’t treat commercial validation as a fundraising box ticking exercise either! Commercial validation is existential. It matters because the cost of fake validation is higher than you realise. Free users will cost you tokens. Pilots will consume your time. Custom workflows will create product debt. And every ‘promising’ prospect who never pays will convince you to build in the wrong direction.

Always start with commercial and work backwards.

Lastly, if you’re fundraising, remember: in AI, many teams can build. In investor conversations, what you’re trying to convincingly explain is that what you are building sits on top of a painful enough problem, with a clear enough buyer ..and a budget that can actually move. Good luck!

Data

IPOs Reloaded.

It looks like Europe’s old IPO backlog is finally clearing, giving way to a new wave of AI IPOs.

According to the latest data data, Europe has recorded 87 IPOs so far in 2026, mainly helped by much lower volatility and stable interest rates. And with Q2 listing volumes up about 30% quarter-over-quarter, there’s finally some reason to be more optimistic about European IPOs!

But before we pop that champaigne, what’s more interesting is not necessarily the number of listings, but the pipeline itself.

There are now 223 European companies with a high probability of going public, down from 373 last year and only 10 now have a greater than 90% probability of listing, compared with 32 just a year ago.

..that’s a nearly 70% drop and the remaining pipeline is also extremely concentrated. Its combined valuation is about €232 billion, with Revolut alone, which is eyeing an IPO later this year, accounting for roughly €200 billion of that!

So yes, the IPO pipeline itself is looking thinner and very concentrated.

Where do the IPO candidates come from? Geographically, the UK dominates, with 37.5% of Europe’s AI IPO candidates based there. France accounts for 18.1%, and Germany 8.3%.

There is, however more to this story and it’s called, you’ve guessed it: AI.

Out of the 223 European IPO candidates, 72 of them are AI companies, meaning roughly one in three companies in the entire pipeline is now AI-related.

..the new AI-native European IPO class is forming!

Community

Female Innovation Index.

Italian spacetech startup ORiS, co-founded by Anna Mauro, raises a €5M pre-seed round led by Earlybird; Hungarian mobility startup Volteum, co-founded by Zsófia Tóth, raises a €2.5M round led by Movens Capital; UK hospitality startup FlairMakers, co-founded by Raluca Epureanu, raises a £500k pre-seed round led by SFC Capital; German female health startup DR. VIVIEN KARL, co-founded by Dr. Vivien Karl and Julia Huhnholz, raises a seven-figure seed round led by two.7 Gruppe; French AI infrastructure startup Pathway, co-founded by Zuzanna Stamirowska and Claire Nouet, raises new seed funding bringing total seed financing to $30M, at a $500M valuation, backed by Id4 ventures; UK cleantech startup Metal Morph, co-founded by Anne-Charlotte Mornington, raises a £700k pre-seed round co-led by Sustainable Ventures and Green Angel Ventures.

Weekly funding coverage from Female Foundry. View the 2026 Edition of the Female Innovation Index.

Events

Tuesday, August 18, Netherlands ➯ Pitch & Pizza | Wednesday, August 19, London ➯ Founders Breakfast Club, Berlin ➯ Co-Founder Matching Night, Zurich ➯ Robot Learning Paper Club | Thursday, August 20, London ➯ Adaptive Demo Night | Friday, August 21, London ➯ All-In Summer Picnic | Saturday, August 22, London ➯ UK AI Agent lab: Gemini Edition.

List your event

Enjoy the week ahead! See you next Sunday.

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Big questions for the AI Age.

The Big Idea

Hot & Scarce.

SpaceX reported its first quarterly results as a public company this week, and buried inside the rocket launches and Starlink projects was something very interesting: Spacex’ AI revenue reached $2.56 billion in Q2, up 247% YoY and more than 3x the previous quarter.

But what became clear was that SpaceX’s new AI business is not only Grok, but increasingly it’s renting out its compute infrastructure.

How much? As much as $1.6 billion of incremental AI infrastructure revenue during the quarter was generated by new cloud-service agreements - that’s part of staggering $14.1 billion worth of cloud-services contracts!

In June alone, Google agreed to rent access to roughly 110,000 Nvidia GPUs, paying SpaceX as much as $920 million every month once the capacity is fully ramped.

And SpaceX is not alone.

AI infrastructure leasing is becoming one of the hottest—and potentially most lucrative—businesses of the AI boom.

Also this week, UK AI cloud infrastructure startup Volta, co-founded by Sofia Gumuzio, exited stealth at a €2.07B valuation and landed an €8.6B compute deal reportedly with Anthropic (see the Index section below).

And to give you a broader context: CoreWeave ended the first quarter with $99.4 billion of contracted revenue backlog, after generating just $2.1 billion of quarterly revenue. Meta alone signed a new $21 billion commitment in March, and Anthropic also signed a multi-year deal in April to get additional compute.

And then there is also European Nebius. Meta agreed to rent $12 billion of dedicated AI capacity from the Nebius over five years in March.

These numbers are staggering especially for, what at its core, is a fairly straightforward business: get hold of the chips, power and data-centre capacity before somebody else does, then rent them out.

But building an AI data centre also takes time:

You need land (reportedly 71% of Americans are currently against AI data centres), transformers, cooling equipment, networking hardware and, most importantly, access to enormous quantities of the latest Nvidia chips.

And that scarcity has now created a remarkable inversion of the traditional cloud model.

For the past two decades, AWS, Microsoft and Google built enormous cloud businesses by renting computing power to everybody else. Now, as the AI race stretches their own infrastructure, the hyperscalers themselves are becoming some of the biggest renters and their enormous contracts help them raise debt and fund yet more GPUs and data centres, turning access to compute into an increasingly financeable asset!

So who are the lucky sellers right now?

Perhaps, rather unlikely AI companies.

IREN (one of the fastest growing hyperscalers) spent most of its life as a Bitcoin miner. CoreWeave also started as a ‘nerdy’ crypto-mining operation. Crusoe originally used stranded natural gas to power Bitcoin mining. .. and now a new kid on the block - SpaceX that also builds rockets.

Of course, there is an obvious catch. Scarcity creates extraordinary economics only for as long as the thing remains scarce. If supply eventually catches demand, today’s $100 billion backlogs could start looking very different.

But that is not what the market looks like today. Will data centres in space be the inflection point that ends that scarcity—or fuel it even more?

GTM Edge

On Proximity.

I spoke with a founder this week who’d been shipping new agents, fixing bugs and integrations at an almost absurd pace, all through a pre-designed and glossy customer relationship UX.

But when I asked how much time he‘d actually spent sitting with his customers lately, the answer was basically - none.

It’s a bit ironic but AI has made customer proximity more important, and I see an entire universe of founders these days that seems to not notice that!

See, when building gets easier, the real customer insight is a huge advantage and it can only be found away from a pre-scheduled call, in the customer’s canteen, an internal Slack thread, and quiet growing frustration no one wants to talk about.

Sure, on your customer calls you will get to know the ‘official workflow’, but only by spending ‘spontaneous’ time with your customer you will suddenly hear about that crucial manual step they forgot to mention, introduce you to the person who secretly approves everything, show you the report nobody reads but everyone is afraid of, the tool they actually may still use besides your product, and the part of the process they hate but have normalised.

So where should you start?

Firstly, remember, proximity is about small, deliberate moments and no customer will ever create them for you.

It is the ten minutes after a demo when you do not pitch, but ask why this problem became urgent now. It is ‘a random coffee’ after a panel at a conference where you ‘also happen to be’ when someone finally says what they would never put in a procurement document. It is at that side dinner you ‘somehow also attend‘ where two company employees describe the same pain in different words. It is visiting a customer when you are ‘already in town’ not to sell - but to understand what happens before and after your product enters the workflow.

See where I’m going? It is all about showing up where your buyers already are: industry events, operator dinners AND creating additional micro-moments: post-demo debriefs, industry roundtables, and quick office visits (get to know they WFH policy first) when there is enough trust.

Once you literally get your foot in the door, remember not to perform thought leadership but rather to hear the unpolished version of the problem.

And the questions you ask and how you use your time during those micro-moments also really matter!

Instead of: Would you use this? Ask: why now? Insead of: How big is this problem? Ask: What happens if this doesn’t get fixed? Who owns this problem internally? What are you using instead? Is this budgeted as software, headcount, services, or pain? What would make this urgent enough to pay for this quarter? Who else needs to believe this is worth changing?

Then follow up. ALWAYS. (Most founders ask questions and disappear!).

After every micro-proximity-moment ALWAYS send a followup note, not with “it was great to see you” but with “It would be good to dive deeper”, “Can I take 10mins of your teams’ time this week to see first-hand how we can get it fixed”?

Get the next meeting micro-proximity-moment in motion. See where I’m going?

Proximity has a beautiful compounding quality when customers see that their input actually shapes the product.

Data

Deep Waters.

Highland Europe and Index Ventures announced new $1B+ funds in recent weeks, and they seem to reflect an interesting difference among some of the $1B+ funds raised in Europe and the U.S. so far this year: it is not necessarily the total amount raised that is the major difference or the number of vehicles, but where the capital is going.

Data on new $1B+ funds raised in Europe and the U.S. shows that, in Europe, growth-stage capital outweighs early-stage capital by almost 10:1. In the U.S., the ratio is closer to 1.5:1. This means that, unlike Europe, across the ocean some of the biggest U.S. funds are backing companies across the full lifecycle—from early experimentation to later-stage expansion.

Europe has long lacked the scale-up capital needed to keep its fastest growing companies from turning to U.S. investors. But now there is another risk: Europe cannot compete at the global scale unless enough fast-growth European startups are well funded also at the top of the funnel!

Community

Moonshot Weekly x European Blockchain Convention.

The lucky winner of the European Blockchain Convention ticket is: Serena (Bu62hd). Congratulations!

Female Innovation Index.

UK AI cloud infrastructure startup Volta, co-founded by Sofia Gumuzio, exits stealth at a €2.07B valuation and lands an €8.6B compute deal reportedly with Anthropic; Also UK legal AI startup Aavalynx, founded by Hanna Roos, raises a £1.5M round led by European Omega Ventures; Swiss robotics startup Exclaim Robotics, co-founded by Helena Oleynikova, comes out of stealth with a €4.29M round led by Cherry Ventures; French customer intelligence startup Shiplog, co-founded by Khushi Mehta, raises a €807k round backed by Kima Ventures and Project Europe; Dutch climatetechstartup Ore Energy, co-founded by Yaiza Gonzalez Garcia, raises a €37.3M Series A round co-led by Plural and HV.

Weekly funding coverage from Female Foundry. View the 2026 Edition of the Female Innovation Index.

Events

Monday, August 10, London ➯ little computers show and tell, Berlin ➯ Startup Social Run | Wednesday, August 12, London ➯ BrainFest | Thursday, August 13, London ➯ The New Standard | Friday, August 14, London ➯ London Office Launch Party, Zurich ➯ AI Enthusiasts Summer BBQ | Saturday, August 15, Zurich ➯ Learning Robot Learning Hackathon.

List your event

Hiring

TechTree ➯ Founder Associate | NOXUS ➯ Founder Associate | THEKER Robotics ➯ GTM & Strategy.

List your job

Have a great week ahead!

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Building for the AI Age.

The Big Idea

The Dance of The Whales.

Index Ventures, the VC firm behind Wiz and Revolut, announced the raise of a fresh $2 billion across three funds this week, pitching its “new focus” on the AI economy.

So far, surprisingly, Index has been less aggressive in AI than many of its large-fund peers.

With Physical Intelligence, Fireworks AI, its late-stage position in Anthropic and a few smaller bets, it has otherwise been relatively quiet during the biggest AI investment cycle in venture history. The new raise now gives Index the firepower to play catch-up, particularly at the early stage.

Why should we care?

Because the catch-up race among the largest VC firms, the whales with $1B+ AUM, is becoming one of the most important dynamics in the market, especially when money flowing into VC funds has been declining for three consecutive years in both Europe and the US—52.4% in 2023, 4.4% in 2024 and 34.7% in 2025—and when the concentration of capital in a small number of AI companies is now being mirrored by a concentration of capital in a small number of VC firms.

As the rest of the market has contracted, many LPs wanted the ‘safer bets’ backing the biggest global VCs (think: a16z, Sequoia, General Catalyst, Lightspeed Venture Partners, in Europe Lakestar, Northzone) that has since 2023 been capturing an increasingly large share of the remaining LP capital.

For context, in the first half of 2026 alone, the top 30 took in more than one-third of all capital raised globally (!) and extraordinary 93.6% of their investments by value this year went into AI companies, up from 57.1% in 2023 (globally, AI deals account for 78.8% of investment value).

That concentration has given the biggest firms exposure to many of the most valuable private AI companies in the world (OpenAI, Antropic, Safe Superintelligence), because they have been able to top up their ownership having the scale to do so across multiple rounds.

For context, twenty-four of the biggest whales have now invested in at least one of the 30 most valuable privately held AI companies. Among them, the ultra-concentrators—(think: Sequoia, Lightspeed and General Catalyst) have each backed ten or more!

The whales have also been increasingly entering at the earliest possible stage. The biggest 30 have participated in 34% of seed investment by value this year, up from just 11.7% in 2023. And it seems that it’s been an easy swim.

The median seed cheque from a whale has reached $9.1 million this year, compared with $3.7 million across seed deals globally and the biggest firms have been using larger funds to price smaller investors out of the most competitive ones.

Will these bets pay off?

Let’s see how it plays out, but what is clear is that the average valuation step-up among companies backed by the top 30 VC firms reached 3.8 times this year, matching the pace seen during the boom of 2021 and 2022. That is 26% above the global average and the widest gap in at least a decade!

What’s also interesting is that as venture returns have declined across the industry, the gap between the best funds and everyone else has also been growing.

Among 2020-vintage funds, top-quartile funds generated IRR twice those of median funds. That is the largest performance gap among vintages and today, the aggregate valuation of companies backed by the top 30 firms stands at approximately $7 trillion—more than double the level in 2023.

All of this leaves one question: will we soon see the dance of the whales pay off in unprecedented ways in VC history, leading to a market-wide consolidation?

With frontier labs needing to sustain their growth, open models entering the scene, and AI companies continuing to absorb an ever-larger share of VC capital, the outcome will not simply be a new generation of dominant tech companies, it will also lead to a new breed of mega VC firms.

GTM Edge

On AI budgets.

I spoke with a founder this week building an AI workflow optimisation product selling to a large enterprise. The product was getting internal traction, but the deal seemed stuck somewhere between the enthusiasm of the first person he’d pitched to and the procurement department.

It quickly became clear during the call that the problem wasn’t the product. It was his lack of clarity about who the buyer actually was!

I spoke with about twenty enterprise buyers over the past three months to better understand where the AI software buyer market is going, and one thing has become clear: AI budgets are now much more spread across teams. For the first couple of years post ChatGPT release AI budgets sat neatly within IT departments, now finance, RevOps, HR, legal, customer success, product, and compliance increasingly have their own AI tools, AI business objectives, and therefore AI budgets.

If your product changes how work gets done, you’re even more likely going to be better off speaking directly with the department that strategically cares for the work to be done.

This means that you should really stop asking: “Who likes the product?” and instead start asking: 1. who owns the workflow (your ‘user’ champion who feels day-to-day pain, loves the product, and engages with it on a regular basis) and 2. the buyer (increasingly the head of the department) who feels the pain strategically, meaning whose promotion, bonus, or performance depends on this going well, who looks good if this works and who has an AI budget and authority.

Before you pitch, prioritise those two accounts. Two other people to think about are: Who can block the deal? Who will be blamed if this fails? (they might be the same person as the budget holder who will ‘look good if it works’).

Then make sure your product, pitch, and functionality to the outcome is tangible, fast to deliver and strategic to them (always prioritising the budget holder).

Data

The April Fool.

It looks like global venture funding accelerated sharply in April, the latest data shows, reaching $56 billion twice (!) the amount invested a year earlier and the third-highest monthly total of the past 12 months. But before we get all too exited, and actually dive deeper into the data, it’s clear the figure was down to a handful of enormous AI deals (Anthropic’s $15 billion round and Bezsos’ Project Prometheus $10 billion), accounting for 45% of all venture capital deployed during the month.

You take them out, and YoY growth drops to 19%. Still formidable.

Diving into the $56 billion, AI companies attracted 66% of global funding ($37 billion), with most of that capital flowing to model developers (think: Ineffable Intelligence, Recursive Superintelligence).

Geographically, still the US has been dominant, capturing 70% of the global total ($39 billion). That’s huge.

Community

Moonshot Weekly x Cambridge Tech Week.

The lucky winners of Cambridge Tech Week tickets are: Alexandre (i6YVB4), Clara (SSB1QT). Congratulations!

Female Innovation Index.

UK startup Intropy, co-founded by Franziska Kirschner, raises a €9.5M seed round led by Felix Capital; UK menstrual health startup DITTO, co-founded by Alice van der Schoot, fetches a €5.2M seed round led by FoodLabs; French dermocosmetics startup Krème, co-founded by Juliette Lailler and Marie Belile, picks up an undisclosed round from Iris Ventures; UK biotech startup BugBiome, co-founded by Alicia Showering, secures a £500k grant from Innovate UK.

Weekly funding coverage from Female Foundry. View the 2026 Edition of the Female Innovation Index.

Events

Monday, August 03, Berlin ➯ Getting European AI Right | Tuesday, August 04, London ➯ London Startups Operator RunClub & Social | Wednesday, August 05, London ➯ Neurotech x Wearables Co-Founder Matching | Thursday, August 06, Berlin ➯ Speed VS Sovereignty | Saturday, August 08, London ➯ Monad Blitz London Hackathon, Amsterdam ➯ HTML Day 2026.

List your event

Hiring

Mater AI ➯ Founders Associate | Tadaweb ➯ Accountant | THEKER Robotics ➯ Talent Sourcer.

List your job

Have a great week ahead!

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Building for the AI Age.

The Big Idea

The Margin Call.

Moonshot AI’s release of Kimi K3 made global headlines on Monday this week as one of the most capable open-weight models in the world, with stunning 2.8 trillion parameters, performing on par with top-tier US AI models on various coding and reasoning tasks, catching the frontier AI community off guard and raising questions about Silicon Valley’s dominance.

The news comes as OpenAI reportedly is delaying its IPO until early 2027, after confidentially filing to go public on June 8 but staying cautious after Spacex’s initial rally and subsequent fall could signal retail investors having less interest in buying.

Will those fears deepen following Kimi’s announcement? And by the time of OpenAI’s rumoured March 2027 IPO, will the path ahead be clear enough to reignite enthusiasm for OpenAI in the fast-moving market?

Unlike most closed Western frontier models, Kimi K3 is being released with open weights, allowing developers to adapt and deploy it without recurring proprietary licence fees. Running it still requires costly chips, servers and electricity, but access to the model itself is no longer controlled by a Western lab.

Why does it matter?

Independent performance benchmarking sites have placed Kimi K3 only slightly behind OpenAI’s flagship GPT-5.6 Sol, yet Kimi costs API customers $3 per million input tokens and $15 per million output tokens—that’s 40% and 50% less than OpenAI Sol respectively. Basically, you get near-flagship performance at a sharply lower price!

And that’s a big deal, because it fundamentally reshuffles the underlying economics of AI.

If developers can access near-frontier performance through an open model, a company may still pay a premium for the very best performance, security and reliability, but it’s less likely when a cheaper alternative exists for ordinary business tasks.

So where do we go from here?

In the near term, this could rapidly compress profit margins. For context, together, OpenAI and Anthropic expect to spend around $65 billion training and operating their models in 2026 alone, and need high subscription and API revenues to justify those spends.

If a temporary lead in benchmark performance is no longer enough to support skyrocketing valuations (for Anthropic $965B and OpenAI $852B), we are facing an uncomfortable question: how durable are the technological moats of frontier AI labs?

OpenAI still has enormous advantages in capital, infrastructure, research talent and global distribution. Closed models may stay ahead on reliability, user experience and specialised performance, but Kimi K3 could actually be good news for startups.

It may make OpenAI accelerate further by integrating with high-growth players across the vertical stack and in core industries such as healthcare, finance, law, and manufacturing—you name it.

Could you be one of them?

GTM Edge

The Free-Tier Trap.

I spoke with a pre-seed founder this week building an AI orchestration layer for internal operations teams. She complained that even though the early product was getting strong traction - prospects were signing up, testing workflows, and giving positive feedback, when it came time to pay, the urgency vanished and she couldn’t close the deals!

Welcome to the free-tier trap.

I see many AI founders fall into it these days. They offer a free trial to reduce friction, let users in, prove value, and build trust, making the product too easy to try… and too easy not to pay for.

Let me explain.

A free tier works only when it creates a clear path to paid usage. It completely fails when it gives away the core value, attracts the wrong users, increases support or infrastructure costs, and delays the most important signal: whether someone will eventually pay. Just to be clear, the question here is not whether free is good or bad. It is whether free is teaching the market to value your product.

And that signal matters from day one, especially for AI founders.

Why? Because free users are rarely cheap users. Compute, model calls, onboarding, support, and your attention all carry a cost! If those same free users don’t begin to pay, you are not building a growth engine, you are merely subsidising your own curiosity!

And so, the key is to let free users understand the value, but make paying users capture that value. Your free product should only get users to the moment where the problem is obvious, the solution is useful, and the next step naturally requires payment.

That means for example that the individual can test, but the business should pay. If you’re a B2B founder, that usually means charging around usage, collaboration, workflow depth, integrations, automation, security, reporting, or production use.

Also, if your product solves a real business problem, you should get comfortable with asking for money early. Payment changes the conversation. It reveals true urgency, budget size, buyer authority, and whether the user is just a fan (and there are plenty of curious window-shoppers out there) or actually a customer.

How do you know who is who? The fastest path is to put a paid offer in front of the customer as soon as they experience the first real value moment: the first saved workflow, qualified lead, resolved ticket, completed analysis, or measurable time saving.

That is the moment you should know and price your product around. What is that moment in your case?

A couple of side notes.

Remember, customers do not really pay for features. They pay because a product saves time, makes money, reduces risk, replaces manual work, improves conversion, or gives them an advantage they can explain internally.

Also, a note on ‘free forever.’ Remember, free should only ever create activation, not full satisfaction. ‘Free forever’ works the opposite. It creates weak urgency and poor qualification. A short trial or a paid pilot is much better for learning because it forces a decision (no decision is also data).

Finally, don’t count on investors getting fooled either. Your big free user base might look like great traction. In reality, only paid customers are the proof that the market is pulling.

Data

Dilution Takes a Cut.

It looks like founders are giving up less equity at every stage of the venture cycle than they were last year and earlier, the latest data finds.

The data covering 6,803 US startup rounds completed between January 2023 and June 2026 show a steady decline in median dilution across seed, Series A, Series B, and Series C rounds.

At seed, median dilution was 21% in both 2023 and 2024, then fell to 20% in 2025 and 18% in 2026. That’s a is 3% decline. So far in 2026, the middle 50% of seed deals involved founders selling between 12% and 22% of their companies.

Series A seems is following almost the same path. Median dilution fell from 21% in 2023 to 20% in 2024, 19% in 2025 and 18% in 2026 with the middle 50% of 2026 Series A rounds ranging from 11%-21% dilution.

But the biggest shift is visible at Series B. Median dilution dropped from 18% in 2023 to 15% in 2024, remained at 15% in 2025 and then has so far fallen to 12% (!) in 2026. That is a 6% decline (!), equivalent to roughly 1/3 of the 2023 level. In 2026, the middle half of Series B deals ranged from 8% to 17% dilution.

By the Series C, median dilution declined from 13% in 2023 to 12% in 2024, 10% in 2025 and 9% in 2026. The middle 50% of 2026 Series C rounds now ranges from just 6% to 13% dilution.

Looking at the macro - In 2023 founders sold a median of 73% of equity in aggregate across the four representative rounds. By 2026, that total had fallen to 57%.

That’s a big, big shift.

Community

Moonshot Weekly x EBC 2026.

Moonshot Weekly is a Community Partner of the 2026 Edition of the European Blockchain Conference, taking place on 16–17 September in Barcelona that will bring together 70,000+ professionals building digital asset markets, infrastructure and policy and we have 2x 750€ FREE tickets to give away to our community!

To enter the FREE ticket draw, submit your interest below. Deadline: 02 August, 10pm BST. T&C of the EBC apply. I will announce the winners here in two weeks and directly via email. Good luck!

Enter The Draw

From Female Innovation Index.

British construction startup Cascade, co-founded by Hannia Zia and Joana Ferreira, raises a $3.5M round led by Andreessen Horowitz Speedrun; Icelandic fashion AI startup Sowilo, co-founded by Heiðrún Ósk Sigfúsdóttir, raises a pre-seed round led by Kría fund; Spanish FoodTech startup Sensesbit, co-founded by Maruxa Quiroga, raises a €1M round from Clave Capital, Eoniq Fund, WindOne Consultores and Paraíso Natural Ventures; British medtech startup Virilitas Labs, co-founded by Simone Immler, raises a £700k+ round from Innovate UK grant and Anglia Capital Group; Stockholm edtech startup imagi, co-founded by Dora Palfi and Beatrice Ionascu, raises a $4.5M round co-led by Brighteye Ventures and Day One Capital; British climatetech startup Furbnow, co-founded by Becky Lane, raises a £2.5M round led by West Midlands Co-Investment Fund and Nesta; German creator economy startup Passionfroot, co-founded by Jennifer Phan, raises a $15M Series A round led by Insight Partners; German AI-first wealth management startup Aimir, co-founded by Lisa Sorg, raises a €12M round led by Headline.

British startup events platform Onstage, co-founded by Issie James, has been acquired by Sifted; British fintech-for-good startup Lightning Reach, founded by Ren Yi Hooi, gets acquired by European Technology Group.

Weekly funding coverage from Female Foundry. Check the 2026 Edition of the Female Innovation Index ➯ here .

Events

Monday, July 27, Munich ➯ Local AI and Robotics Night | Tuesday, July 28, London ➯ First Step: Graduate to AI | Wednesday, July 29, Paris ➯ Prototyping the Future, ➯ AI Builders Night | Thursday, July 30, London ➯ AI After Hours, ➯ DeepTech London Summer Party, Berlin ➯ The Missing Piece in Agentic AI: Payments, Zurich ➯ Forgis Research Night | Saturday, July 31, Barcelona ➯ IT Girls BCN.

List your event

Hiring

Lexroom ➯ Product Manager | Tadaweb ➯ Accountant | THEKER Robotics ➯ Talent Sourcer.

List your job

Have a great rest of the weekend. See you next Sunday.

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Building for the AI Age.

The Big Idea

The Gateway.

Mira Murati’s Thinking Machines released its first long-anticipated model, called Inkling, on Wednesday this week, with data control as one of its key differentiators. The news comes just after I wrote about the rise of open models last week (see: Week 201), and Inkling’s launch adds a new dimension to how value creation in AI models might look in the near future.

Why should we care?

The Inkling launch is interesting on two levels.

First is about data control. With 975 billion total parameters (see more on how open models work ➯ here), Inkling is much smaller than closed frontier models, and Thinking Machines is very open about it. Instead of trying to beat OpenAI or Anthropic on performance benchmarks, it’s offering enterprises models where they can have full control over their proprietary data. And that’s big!

Enterprise adoption is a cake many frontier labs have been agressively trying to get a slice of for months now—remember OpenAI’s and Anthropic’s partnerships with consulting firms earlier this year to ramp up enterprise integration and adoption? However, greater adoption of closed models has also meant greater concerns from enterprises about handing their proprietary data by the labs (even with the strongest disclosures and safeguards in place).

With a trusted open-source model that lives on-prem (and is developed in the west), that pain-point largely disappears!

And so, with OpenAI, Google and Anthropic racing to scoop up more enterprise sales, Thinking Machines’ value proposition is a new gateway - betting on adaptability to unique enterprise needs over raw power.

And that brings me to the second point: its business model.

For a long time, open models have struggled with business models, per-token pricing has been hard to maintain and closed models offering API-based pricing have simply meant too good business! - This must have been one of the reasons why Meta shifted to closed models late last year and OpenAI moved away from open-source after its inception too.

So how do you make money when the weights are public and no one who downloads them is obligated to pay? You provide services. And in the case of Thinking Machines, it means betting on Tinker, its own model-customisation platform offering training, fine-tuning and hosting services for enterprises built around the model.

And it is a big, big bet. Why?

Because you could argue that other AI labs could offer their own ‘Tinker’ platforms leveraging frontier models too - just like that! Also, will temptation to move to much more lucrative (for now) close models will become too big for Thinking Machines? Or perhaps its bet is simply to become the trusted face of U.S.-made open-source AI for enterprises and fill the strategic gap? Time will tell!

But what does all of this mean for startups?

With Thinking Machines entering the scene, reinforcement fine-tuning could soon become as easy as prompt engineering and that would dramatically lower the cost of building specialised AI products. The real advantage for startups would deepen around proprietary data, distribution and deep knowledge of specific workflows.

Perhaps its something worthing thinking about now.

GTM Edge

Discrete Experiments.

I’ve just landed in Barcelona and am here connecting with the tech scene over the next couple of weeks, if you’d like to meet, do get in touch!

I spoke with a founder earlier this week who had been fundraising just to cover his rising AI token costs and fund go-to-market but during our conversation, it quickly became clear that most of the product experiments he had run had no clear path to commercial validation or value!

I get it. Startups are built on experiments. New features, new prompts, new models and to many founders (I mean those that received free compute credits) AI has made all of this feel almost free. Prompts get longer. Agents make more calls. Teams generate more variants, more content, more prototypes and more noise. The price per token may fall while total token consumption rises.

I see many founders easily mistaking subsidised usage for sustainable economics. Credits just hide the bill but they do not remove it!

And so, when it comes to experiments and testing, one rule you can adopt is, instead of asking ‘How cheaply can we run this experiment?’, start asking, ‘What uncertainty are we paying to remove or value to unlock?

Remember, a good experiment should always lead to a commercial decision. Should we build this? Will customers pay? Does this improve retention? Can we deliver it at a healthy margin?

If the answer will not change what you do next, the experiment is probably just activity.

..And the real cost of ‘getting busy’ is far greater than the invoice.

Think about it: experiments not only consume engineering time and customer attention, but also team focus and opportunity. A prototype that takes a day to build can create months of maintenance! A feature that is cheap to generate can still be expensive to support. Ten vague tests can teach you less than one that is carefully designed around a specific business objective.

And so, before you start an experiment, decide what question the experiment must answer. Decide how much time, capital and compute that answer is worth. Decide what success looks like, but also define failure clearly enough that the team can walk away without another round of prompts, models or integrations.

Putting a boundary around the experiment before momentum turns it into a product, a feature or an expensive internal habit is crucial (!) because the goal is not to run more experiments. Speed without discipline only produces faster waste. More prototypes do not automatically create more insight, and more tokens do not guarantee better decisions.

The goal is to buy better information with less time, less capital and fewer tokens.

The best experiment is not the most impressive one. It is the smallest credible test that changes what you do next!

Community

Moonshot Weekly x Cambridge Tech Week.

Moonshot Weekly is a proud Community Partner of Cambridge Tech Week 2026, taking place on 14-18 September in Cambridge that will bring together leading founders, investors and scientists aiming to address some of the biggest problems of our generation, and we have 2x FREE tickets to give away to our community.

To enter the FREE ticket draw, submit your interest below. Deadline: 25 July, 10pm BST. T&C of the Cambridge Tech Week apply. I will announce the winners here in two weeks and directly via email. Good luck!

Enter The Draw

Data

Picky Corporate VCs.

It looks like corporate venture capital is heating up, but the latest data also shows that funding is flowing into only a handful of deals, following the broader pattern we’re seeing across venture capital.

In Q1 2026, corporate investors participated in 1,236 deals, down 18% year over year, while capital deployed reached $220.2 billion—that’s up 416% (!!) from Q4 2025.

65% of capital went into later-stage rounds, and just 3% went to companies raising their first round. Geographically speaking, North America captured 88% of total capital, and software accounted for 80%, driven largely by a small number of AI mega-rounds.

Looking at the context, there are now a record 3,068 active corporate investors globally, with 46 new CVC units launched this year, already up from 32 launched in the previous year.

Yet more than half of CVC teams manage less than $100 million and make six or fewer investments a year. Most still focus on Series A and B, while nearly a quarter now actively use secondaries.

And so, the message here is not that CVC is uniformly stronger. It is that capital is becoming more selective, more concentrated and more focused on later-stage companies, commercial relevance and liquidity.

From the Female Innovation Index:

British event technology startup Forumm, co-founded by Hilary Young, raises a £250k seed round led by an angel investor network; British fashiontech startup Whering, founded by Bianca Rangecroft, raises a $7M seed round co-led by eBay Ventures and Google AI Futures Fund; Spanish medtech startup Time is Brain, co-founded by Alicia Martínez-Piñeiro, raises a €3M pre-Series A round led by Inveready and Clave Capital; British edtech startup Gaia Learning, founded by Kate Heath, raises a £400k equity round led by GC Angels and NESTA; Italian healthtech startup Doctorsa, co-founded by Nadia Neytcheva, raises a €1M round led by PranaVentures; Danish biotech startup Visibuilt, co-founded by Line Kloster Pedersen, raises a €3.34M seed round co-led by EIFO and Unconventional Ventures.

Weekly funding coverage from Female Foundry. View the 2026 Edition of the Female Innovation Index.

Events

Wednesday, July 22, London ➯ Founders vs VCs: Agent Wars, ➯ Women in tech: Engineering dinner, Warsaw ➯ CreativeTech Meetup | Thursday, July 24, Helsinki ➯ AI founders. Sauna. Sea, Berlin ➯ GTM Garage | Friday, July 25, London ➯ Healthcare UX Meetup.

List your event

Hiring

Relay ➯ Senior Operations Account Manager | Toloka ➯ Freelance Forward Deployed AI Engineer | UbiMaster ➯ Business Analyst.

List your job

Enjoy your Sunday. See you next week!

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Building for the AI Age.

The Big Idea

The Open Era.

On Wednesday this week, a popular open source AI developer tool, Ollama, announced a $65 million round and nearly nine million monthly users, joining a number of leading players (think: Cohere, Amazon) in building open-source AI models and tools as enterprise alternatives to closed models.

Why does it matter?

Because if the open-source trend continues—and everything points to the fact that it will!—the reshuffling of the AI stack will change how startups build.

For the past two years, the market rewarded scale above everything else. Bigger models, larger context windows, more compute, more spending. But now that’s changing.

In May, it was reported that Uber burned through its annual AI budget in four months, as the company started to ramp up its AI strategy. The news was met with disbelief because most AI work does not require the most expensive model on the market. A document summary or a classification task does not need frontier-level reasoning and for a company (or a startup) processing billions of requests, using the wrong model is not a technical mistake anymore - it is a business-model mistake.

And so, as more and more companies are now looking at long-term economics of adopting AI, open-source AI is having its moment, and the Chinese open models, that are often 60% to 90% cheaper than those of OpenAI and Anthropic, seem to be leading the pack (see: Week 176).

Just in the past four weeks, Airbnb dropped ChatGPT over Alibaba’s Qwen for parts of its customer-service stack and DoorDash adopted Chinese open models in their internal engineering pipelines, all to protect their margins. And they are not alone.

What’s striking is that around 41% of AI models downloaded last year were developed in China and the share of Chinese models’ share of tokens used by US companies rose from 11% to as high as 46%. Meanwhile, since January, the combined share of OpenAI, Anthropic and Google has fallen from 55% to 33% (!).

What’s driving the demand?

The attractive part is that open models can be run locally, adapted to specific workflows and deployed without paying a provider on every request. They still require infrastructure and engineering, but they give companies more control over the one thing AI startups can no longer afford to treat casually: unit economics.

And cost is only half the story. The other half is access.

Frontier closed models not only come with pricing risk but also political risk. Recent restrictions on Mythos and Fable (see: Week 197) show that a startup can build around a system and still lose access because of policy. By contrast, once open weights are released, open models are much harder to take back. That continuity makes open them very attractive!

And that’s not the end of the story.

Over the past couple of weeks there’s been a growing number of reports of US lawmakers investigating American companies using Chinese models (including Airbnb and Cursor). The concern has been that cheap, Chinese models could make US companies dependent on Chinese technology, with risks around security, censorship and supply chains.

But the irony is that restricting Chinese models could also raise costs for the very startups the US wants to keep competitive!

So is China the ultimate winner? Not so easy.

Last week, it was reported that Chinese authorities had been looking to curb overseas access to future models from Alibaba and ByteDance. No final decision has been announced as of today, but the debate is telling: China built global influence through openness and it is now asking whether the most capable models have become too strategically important to share.

The story is unfolding, but one thing is clear: as models become cheaper, more open and easier to switch, power will shift from model providers to the companies that control data, distribution and customer relationships. Will geopolitics accelerate that shift—or fracture it open?

GTM Edge

In the Loop.

I spoke with a founder this week who told me she was trying to cut back on weekly updates. Not because the company was slowing down, but because keeping everyone in the loop had started to feel like a second job.

She was collecting updates from her eight-person team, rewriting context, summarising decisions and repeating the same information across Slack, meetings and one-to-ones. By the time the update was ready, half of it was already old! (What a nightmare). Of course, she would rather spend that time with customers, hiring or working on the product.

And fair enough! Time is your most valuable asset in the age of AI. But when you’re building, the answer is not fewer updates. It’s making updates part of the work itself!

Let me explain.

The real problem I see with many founders these days is that communication becomes a separate process, often owned almost entirely by the founder who ‘carries the vision’. That is a lot of time being lost. And so, a better approach is to make progress visible as it happens.

Every deal that closes should be shared in the same channel. So should the ones that are lost. The same goes for product launches, customer feedback, important decisions and lessons the wider team can use.

The beautiful thing about live updates is that onboarding and tracking progress become easier too. You can even add an agent that turns those updates into a weekly team summary.

The person who closed the deal gets recognised publicly. And when something does not work, the team can learn from it while the context is still fresh.

It definitely does not need to wait for an all-hands. All-hands meetings can then be used for strategic steps and steering the ship, not for routine updates.

Where do you start? Create one public channel for meaningful company movement and make the format painfully simple: what happened, why it matters and what happens next.

A deal closes, it gets posted. A customer churns, the reason is shared. A feature ships, the team sees who it was built for. A major decision is made, the context is written down once.

So don’t wait until Friday to reconstruct the entire week from scattered messages and half-remembered conversations. And don’t turn internal communication into a polished performance.

Your job is to set the rhythm, not to write every line.

Because when information flows as the work happens, the weekly update becomes a recap rather than a reconstruction. It takes less time, creates more ownership and gives the whole company a clearer sense of momentum!

Remember, it may seem to you that your startup moves quickly, but that speed is hard to feel for the rest of the team when progress is trapped inside individual meetings, private calls and emails.

So share the wins while they’re fresh. Share the losses while they’re useful. Make the company visible to itself.

Community

Measuring Resilience with GlycanAge.

On Thursday in London, the GlycanAge team (AI Visionaries Cohort I), is running a FREE masterclass on immune ageing, that will explore how immune resilience can be measured and tracked to reflect your body’s capacity to withstand and adapt to real-world interventions.

​The evening will be set in one of London’s most beautiful wellness spaces, and will be a great opportunity to reset and gain a fresh perspective on thinking about health.

Grab your spot

Data

The Serial Advantage.

It looks like Europe’s tighter funding market is giving repeat founders a clear edge, the latest data shows.

In 2026, repeat founders have been raising a median €4.3M per deal, 72% more than the €2.5M secured by first-time founders.

And when you look closer, the gap is even wider in early-stage and AI deals. Looking at the medians, experienced European AI founders raise 92% more capital than first-timers, while repeat founders reach a median post-money valuation of €17.2M, compared with €9.4M for first-time founders (!).

For context, repeat founders now account for 60% of European unicorns founded over the past eight years and represent $2.2T in combined enterprise value.

And so, as European investors concentrate capital in even fewer deals this year, a founder’s track record clearly carries a lot of weight!

From the Female Innovation Index

UK medtech startup Respiro Diagnostics, co-founded by Alison Quinn, raises a £1M pre-seed round co-led by Zinc Venture Capital and SFC Capital; British AI startup Worldmodeldata, founded by Rhea Loucas, raises a £7M seed round led by Iona Star Capital; Swiss photonics startup Aylight, co-founded by Bahareh Marzban, raises a €4.5M pre-seed round co-led by Elaia and Swisscom Ventures; French deeptech / diagnostics startup En Carta Diagnostics, co-founded by Margot Karlikow, raises a €5M round led by Blue Forest Ventures; German-Portuguese deeptech startup alqem, co-founded by Hanh Nguyen, raises a €8M pre-seed round co-led by UVC Partners and Union Square Ventures; French HRTech startup Skello, co-founded by Quitterie Mathelin-Moreaux and Emmanuelle Fauchier-Magnan, raises a €200M growth round led by Bridgepoint.

Weekly funding coverage from Female Foundry. View the 2026 Edition of the Female Innovation Index ➯ here.

Events

Tuesday, July 15, Paris ➯ CLMX SUMMER CLUB | Wednesday, July 16, London ➯ Measuring Resilience: A Masterclass on Immune Ageing | Thursday, July 16, Paris ➯ Anti-networking event for Founders and AI enthusiasts, Berlin ➯ GTM Garage | Friday, July 17, London ➯ Reimagine Health.

List your event

Hiring

doinstruct ➯ Enterprise Account Executive | WeeFin ➯ Senior Sales Executive | Relay ➯ Chief of Staff to Ops and Engineering.

List your job

Have a great week ahead! And see you next Sunday!

Agata

Written by Agata Leliwa Nowicka, the Founding Partner of Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Moonshot Weekly - Building for the AI Age.

📣 Announcement

Female Foundry Enters the Moonshot Era.

After 200 weeks, Female Foundry is joining Moonshot Weekly — a new publication for visionary founders, capital, and European innovation.

For the past years, I’ve been writing about the extraordinary era we are living through: a moment in history where visionary founders have more tools, reach, and opportunity than ever before to build companies that don’t just scale, but shape markets, industries, and the future itself.

During this time, I’ve been meeting extraordinary founders, scientists, operators, investors, movers, and shakers building what comes next.

Today, I’m excited to finally make it public:

Female Foundry has joined Moonshot Weekly to double down on its original promise — shaping the conversation around innovation, frontier thinking, and real-world impact of AI.

Why Moonshot?

Because today demands big ideas and ‘a moonshot’ is to build at the edge of what feels possible.

With AI Visionaries, I saw what happens when ambitious founders leverage frontier technologies, ask hard questions, and carry the conviction that the future is something we build — not something we wait for.

Moonshot Weekly is built on that same belief.

It is about taking that ‘big idea’ and asking what’s needed to turn it into companies, capital and systems with huge real-world impact.

What does it mean for the Newsletter?

Moonshot Weekly will have a weekly structure, built around the pace of innovation: the ideas, founders, funding, data, events, and opportunities shaping the ecosystem.

I speak with 20+ founders every week, and while it has never been easier to start a company, the bar for competition has never been higher.

In this section, I’ll share insights from those conversations, along with practical steps and observations on go-to-market, growth, positioning, and building momentum.

Above all, this space has always been about community and that will continue!

With more than 7,000 VCs, founders, and innovators tuning in each week, I want to create even more opportunities for us to meet, exchange ideas, and build meaningful connections across the ecosystem.

As you know, I love data.

This section will continue to bring together the latest reports, analysis, and funding insights to help founders understand the market context behind what’s happening — and what it means for them.

This section is where we’ll share curated hiring opportunities and upcoming events from across the ecosystem — from startups growing their teams to gatherings worth having on your radar.

Thanks to all of you who’ve been on this journey.

.. and welcome to the Moonshot Era 🌚!

Agata

Written by Agata Leliwa Nowicka, the Founding Partner at Moonshot Capital, serial entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and the Wharton School.

Are you building at for the intelligence age? Let me know.

Suggestions? Drop me an email.

Check moonshotcap.vc for more information. View other Moonshot Weekly articles.

Female Foundry - the Future of Venture is Here.

This is a special Edition of the Female Foundry Newsletter. Read on to see what’s next for the Female Foundry Newsletter after six years of driving impact.

Spotlight

My Letter to You.

‘Hello World’!

It’s Sunday. Thank you for tuning in, perhaps for the second time, perhaps for the 200th?

When I started Female Foundry six years ago, during one sunny but blisteringly cold London weekend in February 2020, I saw it as my way of giving back to the venture ecosystem I had been building in — first as a founder of two tech businesses and now as a VC investor.

As one of the very few women in Europe with such a dual profile (still!), I saw it as a new opportunity for me to stay curious and help hungry and ambitious founders succeed faster.

What followed was this beautiful adventure.

Honestly, I never intended Female Foundry to become a fund — and so many of you wanted me for years! Or a media platform?

It would be just too pragmatic. Too transactional. Too calculated for something that was deeply personal.

See, my vision for Female Foundry was much more simple. For it stay free. To stay fluid. To stay curious.

When I launched Female Foundry in 2020, we were one of the very first to focus on VC-backable female-founded startups in Europe and putting the word ‘Female’ in its title was a big deal at that time. Ah, we’ve come a long way from that!

Since then, Female Foundry has outgrown its newsletter and been on a trailblazing journey of creating female “firsts.”

The first pan-European analysis of female founder funding in Europe. Started in 2022, the Female Innovation Index has since grown into the biggest annual analysis of female innovation and funding in Europe, attracting tens of thousands of readers each year and bringing together national venture capital and private equity associations from 20 European countries as partners.

-This year’s edition was the biggest yet, with over 800 founders signing up to join the 2026 Edition Launch in March. That’s eight-fold growth post-first edition!

Then, there was the first female-focused AI accelerator in Europe. In 2024, just a few months after ChatGPT launched, we built AI Visionaries in partnership with Google to accelerate the growth of some of the most ambitious deeptech and AI-first founders building with AI.

We knew AI would be big — sounds so obvious now but at the time we were navigating truly uncharted waters with the Google team and were still willing to take the plunge!

Throughout this time we’ve partnered with some of the leading universities, research institutes, hyperscalers and the biggest publishers, to create a new, positive, innovation-led narrative for female entrepreneurship in Europe.

And what a ride it has been!

Female Foundry: End of Year Media Campaign 2025

Throughout the past six years, the landing pad for Female Foundry has been its weekly newsletter. Every week — come rain or shine — in your inbox.

Together, we’ve gone through so many new dimensions of startup building!

From navigating remote-only building in 2020, some of the hottest VC capital deployment years in 2021–2022, the VC crash of 2022–2023, and now the AI-building era from 2023 onwards!

Today, over 7,500 VCs and founders read the Female Foundry newsletter on a weekly basis.

And today,

I’m excited to share that we are entering a new era for Female Foundry, keeping the spirit of the community we’ve built here and staying true to its values of building at the frontier.

The new edition of the Weekly Newsletter will launch here next week, and Female Foundry will be part of it.

Thanks for those 200 editions, and excited for the 200 more to come!

Stay tuned. Stay curious.

.. see you next Sunday.

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

I hope you’re somehow beating the European heatwave and keeping cool!

As we edge towards the 200th edition of the Female Foundry Newsletter — I can’t believe I’m writing this — next week I’ll be unveiling the next phase of this space. So keep an eye out next week for the Female Foundry Newsletter celebration, and for updates on where we go next.

For now, let’s dive into this week’s reflections on the world of AI, startup building, and investments!

In the news

Spanish legaltech startup Kalipso, co-founded by Virginia Debernardi, raises a $3.2M round led by Varsity; Italian procurement intelligence startup Soource, co-founded by Silvia Chiarot and Serena Galli, raises a €3M seed round led by Vertis; French medtech startup Tissium, co-founded by Maria Pereira, raises a €30M Series D-2 equity round led by an undisclosed US investor; Danish aerospace startup Acodyne, co-founded by Jasmina Pless, raises a €2.5M pre-seed round co-led by Gungnir Capital and PSV Hafnium; British foodtech startup WNWN Food Labs, co-founded by Ahrum Pak, raises a £4.5M Series A round led by PeakBridge.

London-based Seedcamp co-founded by Reshma Sohoni closes a €165M fund which will see it expanding its presence in the United States.

Spotlight

Vertical Gravity.

It was reported this week that Google has capped Meta’s use of Gemini to ease pressure on its strained AI capacity, raising a few pairs of eyebrows and yet offering another sign that tech giants are struggling to manage surging demand for advanced models against limited computing power.

As AI moves from experimentation to daily use, more and more big tech players are now getting vertically integrated, wanting to control not just the model, but the full supply chain underneath it: chips, cloud capacity, software, power and access.

Within just the past four weeks, xAI (Musk is a master of vertical integration) turned Colossus into a full-stack AI infrastructure bet and OpenAI unveiled Jalapeño, its first custom AI processor, developed with Broadcom. By ‘owning’ Jalapeño, OpenAI is reportedly aiming to cut infrastructure costs by roughly 50% compared with traditional GPU-based systems.

Why should we care?

Taking as an example Jalapeño, if it works at scale, it will not just improve margins. It will change how OpenAI can price, launch and expand its products. Cheaper inference will mean more room to offer advanced models, more compute-intensive tools and more aggressive pricing without letting costs run away.

With more hyperscalers now looking to cut costs by vertically integrating, surely this must be great news for startups running on top of big LLMs?

Not so easy. Because vertical integration cuts both ways.

As a consequence, for example, the easy version of an ‘AI startup’ — take an API, wrap it in a nice UX, sell it as a product — can get even more squeezed and difficult to defend. The more the giants control the full stack, the more they can copy, bundle or underprice thin application layers… and if economics do not make sense, there is no business case.

GTM Hustle

Mini Markets.

Another busy week. Four countries in two weeks.

Typically, I write here about big ideas: building category-defining companies and thinking global from day one. But this week, I spoke with a great founder who’s been obsessed with his ultra-niche market. And it made me think.

For a long time, software only made sense if the market was huge. The cost of building was high, the GTM team required large, and distribution expensive.

These days, however, I’m meeting more and more founders building for niche, specialised markets that are finally economically viable. Suddenly, I’m seeing a new influx of ideas that use to be ‘too niche’ that now start to make economic sense.

Apart from cutting building costs with AI, the strength of starting small today is that you can know your customer better than anyone else. You can understand their workflow, language, constraints, habits, politics, and buying behaviour at a level a larger company cannot. You can build with unmatched precision. You can sell through trust instead of noise. And, even more interestingly, you can build a community before anyone else even recognises the group as a ‘market’. And that’s a huge moat.

Let me dive deeper.

Think about it, if you can build well within a ‘tiny market’ - a concentrated group of people with shared pain, shared workflows, shared language, and shared trust channels - this is a much smarter starting point than a massive theoretical market where nobody feels the problem deeply enough to care.

That, by the way, doesn’t mean building a tiny tool for a tiny problem. A tool will get copied. A tool will get bundled. A tool will get absorbed by a bigger platform the moment the use case becomes obvious (see above). The opportunity now is to build around a community so well, that the product becomes the natural system around which their work starts to organise. ..do you follow?

Because this is a real edge of ‘a mini market’. It is not being small for the sake of it. It is being small for concentration.

Starting small lets you move faster because feedback is tighter. It lets you build defensibility through proximity, not just features. It lets you distribute more efficiently because the buyers know each other. Oh yes, they often do! And it lets you then expand from a position of trust.

Will every mini-market startup have the profile to become a unicorn? Absolutely not. Probably not even close.

But that doesn’t mean a ‘tiny market’ cannot be your GTM strategy on the way to building a global company! It can be the place where your value proposition starts compounding.

(See? Somehow I always land on becoming global).

The old question was whether the market was big enough. Now, perhaps another question worth asking is - is the community you can build around the problem concentrated enough, the problem painful enough, and the market reachable enough for you to win?

Community

Female Foundry x GITEX AI Europe.

Female Foundry is a partner of Gitex AI Europe, the annual expo bringing together entrepreneurs, investors, and thinkers, pushing real-world tech solutions forward, taking place on 30 June–01 July in Berlin and I have one complimentary ticket for our Newsletter Community.

I’m going to give it away to whoever responses first to this newsletter. .. Ready. Steady. Go!

Analysis

The Story of Two Halves.

The 2021 venture boom is finally showing up in VC fund returns, and the view is not that great for GPs and LPs that closed funds between 2018 and 2021, the latest data finds.

In 2021, a lot of capital was deployed at prices that assumed everything would go right: growth would keep compounding, exit markets would stay open, and private markups would eventually turn into cash.

According to the latest data, funds raised from 2012 to 2016 still look solid, with median IRRs mostly in the mid-teens. The 2012 vintage is around 18%, with top-decile funds above 40%. Even 2016 is still near 16%, with the best funds in the mid-30s.

However, from 2018 onward, the story starts to look very different. Median IRR falls to 12% for funds raised in 2018, 9% for 2019, 7% for 2020, and only 4% for 2021.

And the performance gap between good and bad funds raised from 2021 onward is also widening. Bottom-decile funds are around -10% for 2021, -15% for 2023, and nearly -25% for 2024. At the same time, some top-decile funds are still above 30%, that’s a 40–60 point spread between winners and laggards (!).

The VC market is clearly no longer rewarding exposure alone. It is now rewarding precision: entry price, ownership, reserves, DPI, and timing. Funds that treated 2021 as normal are now paying for it. ..Funds that treated it as expensive still have room to compound.

Hiring

This week hiring:

Relay ➯ Chief of Staff Associate | doinstruct ➯ Customer Operations Associate | Cambridge GaN Devices Ltd ➯ Management Accountant.

List your job

Events

This week

Monday, June 29, London ➯ Building the GTM Engine | Tuesday, June 30, Paris ➯ Where AI Engineering Meets Product Design | Thursday, July 03, Berlin ➯ The future of AI in Bioprocesses, Munich ➯ Building with GenAI — An engineering evening, Manchester ➯ CONEXEN Manchester July | Saturday, July 04, London ➯ The Startup Stroll.

List your event

Stay calm. Stay cool. See you next Sunday!

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

Another very strong fundraising week by the European female founders!

Italian hospitality startup Bestie Bite, co-founded by Carlotta Robbe Di Lorenzo and Caterina Vertefeuille, raises a €1.5M round led by the Grassi family through their family holding company G&G; Italian edtech startup Sirius Game, co-founded by Laura Cesaro, Fabiana Gubitosa, and Miriam Torregrossa, raises €1.3M round led by Cassa Depositi e Prestiti; Belgian agtech startup Rainbow Crops, co-founded by Hilde Nelissen, raises €9.7M seed round led by LIFTT together with LIFTT EuroInvest; Spanish biotech startup Gate2Brain, co-founded by Meritxell Teixidó, raises €7M strategic investment from Shilpa Biocare; British company Go Henry, founded by Louise Hill, gets acquired by Barclays; British AI healthtech startup Frontier Health, founded by Rachel Finegold, raises a £12M seed round led by Atomico; British foodtech startup Spoon Cereals, co-founded by Annie Morris, raises a £1.5M seed round led by Finance Yorkshire; Finnish biotech startup Aplagon, co-founded by Riitta Lassila, raises a €4.8M round led by Fåhraeus Startup and Growth and the European Innovation Council Fund; German healthtech / medtech software startup Alpha Sophia, co-founded by Isabel Wellbery, raises a $1M seed round from P3A.

All female-GP VC firm Thena Capital, co-founded by Tatum Getty, Esther Reynal de St Michel Richardot and Pamela Walker Geddes, raises £45m for debut fund to back medtech startups at the Seed to Series A stage.

Spotlight

More and More.

The dust hasn’t even started to settle on SpaceX’s blockbuster IPO last week, and the company captured headlines again on Tuesday with its $60 billion acquisition of Cursor — this year’s largest startup M&A deal. The news sent SpaceX’s IPO price up 16%, pushing it past Amazon and Microsoft by market cap, and made it the fourth most valuable company in the U.S.

The Cursor deal will bolster SpaceX’s efforts to compete with rivals like Anthropic and OpenAI, which offer popular coding tools. It also sent a strong signal of the possible start of a SpaceX’s M&A spree, and of just how much hyperscalers are willing to pay to stay competitive right now.

.. and that’s why we should care.

Looking at the past couple of weeks alone, Salesforce’s $3.6 billion acquisition of Fin valued the AI customer-service startup at about 2.8x its last public valuation (!).Last month, SAP purchased Prior Labs for a reported $1.16 billion. What do all these deals have in common? Yep. They are huge. Yep. They all happened in the past couple of months, and lastly: they are all acquisitions of companies showing accelerated growth.

According to reported data, at the point of purchase, Fin generated roughly $100 million in annual AI-agent revenue, had 350% growth, and achieved a 76% automated resolution rate. At the time of acquisition, Cursor was only four years old, and operated in one of AI’s fastest-monetising categories: coding tools.

And so, buyers are now paying for acceleration, not just current revenue, valuing revenue velocity, automation rates, enterprise adoption, and scarce technical talent. ..We are so back in the growth game!

AI Hustle

The Human OS.

Another busy week. On Wednesday, I spoke with a founder whose eight-person compliance startup had hit an AI adoption wall. The AI tools were there. The freedom to experiment was there. But nothing was compounding.

That is the trap I’ve been seeing when speaking about scaling with a few AI startups lately. The founding team moves quickly, assuming AI adoption will happen naturally and as the team scales, people try new tools, randomly automate small tasks, write faster, research faster, ship faster. But individual productivity does not automatically become company’s strategic advantage.

Why is that?

If you want to make your team operate faster, at any given time you need as a team a clear answer to three questions: what are we trying to make faster, what are we not willing to automate, and where does human judgment still own the final call?

Without that, AI becomes scattered. One person will use it for sales outreach. Another for product specs. Another for customer research. Everyone gets a little faster, but the company does not actually learn faster. And it is learning that closes the compounding loop.

The best AI founders I’ve met in the past had one rule: they treated AI adoption like culture: something that compounds only when it is visible, repeatable, and safe to share. If people are quietly using tools because they are unsure what is allowed, the startup loses the learning. If people are afraid that showing productivity gains will simply raise expectations, they hide the upside. And ..if workflows are messy, AI just makes the mess move faster. That is how you build organisational leverage.

And so, before you build (or buy) that next internal AI optimisation tool, at your next ‘All Hands’ meeting, discuss one painful workflow that could be strategic across several functions (sales-marketing-product?). This could be proposal writing. Customer support triage. User interviews. Whatever creates the most drag. Then measure the before and after: time saved, quality improved, cycle time reduced, decision speed increased.

Also, remember that every new person in your startup will bring their own AI habits, risk tolerance, and assumptions about quality. If you do not define the rules early, the team will create ten different versions of ‘how we use AI’! So make it explicit: what data can go into tools, what outputs need human review, what good prompting looks like, what tasks should be automated first, and what work should stay human because it depends on trust, taste, context, judgment .. and over time creates culture you want to build.

Community

Scaleup Soirée — Staying Competitive in the AI Era.

It was great to bring together fifteen Series A–Series B founders on Tuesday this week for an exclusive dinner in London in collaboration with the fantastic team at ➯ Rathbones.

I learned so much about how scaleup founders think about growth right now and what it takes to compete in the AI era. The dinner was held under the Chatham House Rule, so I can only share the pictures. Until next time!

Analysis

Not Global.

It looks like the AI funding boom might not be as global as it may feel like.

In numbers, so far in 2026, U.S. startups have captured nearly 80% of global seed-through-growth funding, the recent data shows - a sharp jump from the pre-AI boom, when the U.S. typically took less than half of all capital deployed. The gap is even wider when zooming into AI companies: nearly 88% of global AI startup funding, or $319 billion, has gone to U.S.-headquartered companies. The concentration is being driven heavily by mega-rounds, especially OpenAI and Anthropic.

Outside the U.S., only a few major markets are showing real momentum. China has already raised more than $33 billion in 2026, surpassing its full-year 2025 total. The U.K. has reached $16.5 billion so far this year, close to its $19.5 billion total for all of 2025, with AI and fintech leading the market. Other mid-sized venture markets are seeing funding levels this year that are on par or moderately higher YoY. In Europe, this includes France, Spain and Germany.

Hiring

This week hiring:

doinstruct ➯ Enterprise Account Executive | WeeFin ➯ Senior Sales Executive | Relay ➯ Chief of Staff to Ops and Engineering.

List your job

Events

This week

Tuesday, June 23, London ➯ Opening Reception: London Climate Action Week, Paris ➯ WORLD MODELS — A Paper Workshop | Wednesday, June 24, Amsterdam ➯ Future of Product #4: Agentic AI, Berlin ➯ In the /loop - An evening with AI-pilled builders | Thursday, June 25, London ➯ AI x Bio Summer Drinks | Saturday, June 26, London ➯ {Tech: Europe} London AI Hackathon.

List your event

It’s been a good week. Enjoy your weekend and see you next Sunday!

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

It’s been fantastic a couple of weeks for female entrepreneurs in Europe!

Estonian medtech startup Nanordica Medical, co-founded by Olesja Bondarenko and Anna-Liisa Kubo raises €1.6M round led by 2C Ventures; French surgical AI startup Uncovr, co-founded by Ines Iraki, raises $7M seed round led by Index Ventures; Hungarian fleet management startup Volteum, co-founded by Zsófia Tóth, raises €2.5M round led by Movens Capital; UK healthtech startup 01Health, founded by Sonia Szamocki, raises €12.9M Series A round led by Gresham House Ventures; Spanish AI robotics startup Theker, co-founded by Carla Gómez Cano, raises $85M Series A round led by CRV; UK voice AI infrastructure startup AethexAI, founded by Mariama Diallo and Ayooluwa Odemuyiwa, raises $3M pre-seed round led by 4DX Ventures; Dutch cultured wood startup New Dawn Bio, co-founded by Kianti Figler, raises €2.1M pre-seed round led by CapitalT; UK deeptech AI startup Apoha, co-founded by Anshika Srivastava, raises $36M round led by Singular; UK industrial AI startup Gigaton, co-founded by Buffy Price, raises $26M Series A round led by Plural; UK AI training data startup Poindexter Labs, founded by Jocelyn D’Arcy, raises £2M seed round led by Episode 1; UK healthtech startup Rem3dy Health, founded by Melissa Snover, raises €16M round led by a group of global strategic investors including Suntory and Apollo Hospitals; UK medtech startup Joyvié Health, founded by Zoe Robson, raises €897K pre-seed round backed by a number of angel investors and syndicates; French quantum computing startup Quobly, co-founded by Maud Vinet, raises €115M Series A round led by Bpifrance, SEALSQ and STMicroelectronics; Norwegian cloud infrastructure AI startup Cloudgeni, co-founded by Iuliia Petryshyn Thuen, raises €858K round led by byFounders; UK baby and children’s food startup Lune & Wild, co-founded by Lara Rodgers and Nadia Simonds, raises €2.3M Series A round led by Guinness Ventures.

Spotlight

Asking For Permission.

SpaceX made history on Friday with the biggest stock market debut after nearly two and a half decades as a private company. Long-anticipated public trading kicked off at a starting share price of $150, which quickly jumped by a double digit percentage and sent the company’s valuation above $2tn, where it remained through market close and now. The IPO made Elon Musk, the world’s first trillionaire and opened up the pathway for Anthropic’s and OpenAI’s IPOs (See last week).

Also on Friday, the US government ordered Anthropic to immediately shut down the access to two of its most powerful AI models — Fable 5 and Mythos 5 — citing national security concerns. Anthropic announced on X that it followed the order, however, it also made clear it thinks the government got it wrong.

Why does it matter?

For the past two years, the industry has been obsessed with one question: who has the best model? Better benchmarks, bigger context windows, faster reasoning, cheaper inference, more capable agents. Every launch has been a scoreboard. OpenAI beats Anthropic on this test. Anthropic beats Google on that one. DeepSeek changes the cost curve. Meta opens another model.

But Friday’s shutdown order against Anthropic points to a different future. The U.S. government has changed the assumption that frontier AI companies get to decide when their own systems are safe enough to release and that is a much bigger shift. And so, the next phase of AI may not be defined by who can build the most capable model. It may be defined by who can survive the politics of having one.

And the irony is that Anthropic has built its entire brand around being the ‘responsible lab’. The one that takes safety seriously. The one that talks about responsible deployment, controlled access, red-teaming and frontier risk. Mythos was still kept restricted. Fable was positioned as the safer commercial version. In the old AI race, that should have been enough. In the new one, it became part of the problem!

Let me explain my thinking: Once a company spends years telling the world that its models are powerful enough to require special handling, governments may start treating those models as powerful enough to require political control.

That is the new tension. The industry wants to prove that models are becoming more capable, more autonomous and more economically essential. But the more successful that argument becomes, the less those models look like normal software. They start looking like infrastructure. Strategic assets. Systems with national security implications. And governments do not leave strategic assets to company policy pages ..!

And so, the frontier is no longer just technical. It is political.

For application-layer startups, this means the platform risk in AI is no longer just pricing, latency or a model provider launching a competing feature. It is the possibility that the model underneath your company becomes a regulated asset overnight.

For investors, questions about regulatory resilience have just become even more real. Can a startup switch models? Can it explain its safety architecture? Can it operate if a certain capability is restricted? Can it sell into enterprise and government environments without becoming a compliance nightmare?

The first era of generative AI rewarded speed. The next era may reward companies that can move fast without becoming impossible to approve, insure, govern or trust.

AI Hustle

The Scars.

I spoke with a founder this week who’s been building an AI workflow tool for mid-market finance teams. The product was working in pilots, but sales calls were flat. Some prospects immediately understood the value. Others nodded through the demo, asked for a follow-up, and disappeared.

His first instinct was to fix the obvious: tighten the messaging on the website and the demo. Make the onboarding shorter, the before-and-after examples more concrete and show use cases with the sharper ROI. That’s where I see most founders go first, because it feels controllable. - Improve the story, reduce the friction, make the product easier to understand.

But in reality the actual issue might be very different!

What quickly became clear to me was that his customers were not confused because they had never thought about the problem. They were confused because they had already tried to solve it!

One finance lead might have been using ChatGPT to summarise budget. Another had built a messy workflow across Excel, and an internal chatbot. Someone else had tried feeding past board packs into a model to recreate the structure, but got nervous when the numbers and language started drifting.

Those were not passive buyers waiting to be educated. They were people with scars.

And so, by the time they entered the sales conversation, they were not simply asking what the founder’s product did. They were comparing it against their own failed experiments!

Could it handle messy spreadsheet inputs? Could it stay faithful to company-specific language? Could it explain variance without hallucinating? Could it save time without creating a review burden? Could it fit into the reporting process without asking the finance team to change how they already work?

That’s the crux of the issue I often see founders miss altogether.

I see founders still sell from the product outward. Here is the workflow. Here is the automation. Here is the value. But today customers buy from the experiment inward. They also leverage AI to solve problems that they couldn’t solve before, they also aim to make their work easy.

That changes the sales motion.

And so, if you sell to enterprise today, just assume that the buyer has already done part of the discovery, don’t start with ‘what is your current process?’ start with: ‘what have you already tried with AI?’ That’s the real start of the sales process!

In their answer (and it’s also your job to unearth it!) is almost everything you need:

1.You should learn what problem is painful enough for them to experiment with before buying (experimentation in enterprise has budget assigned to it!). 2. You learn what workflow already has internal momentum (where they are in the journey and where you can capture them). 3. You learn what failure modes they have seen. (yep, are you one of them?) 4. You learn what language they use to describe value. (here is the warm-up to the pricing conversation) 5. You learn what they trust, what they reject, and where the product needs to be more opinionated. (hold pinning down your world thesis in front of them until this is clear).

Final note:

If a prospect has already hacked together a workaround, do not rush past it to show your demo. Slow down. Ask what triggered the experiment. Ask what they used. Ask where it almost worked. Ask where it broke. Ask who saw the output. Ask what would have made them trust it enough to use it again.

That failed experiment may be as well your best follow-up sales asset yet (!)

Community

AI Hustle Live #5.

I’ve been travelling for the past four weeks and it’s great to be back in London for the next few weeks. loved learning about what you’re building and giving you hands-on GTM feedback during our last AI Hustle live session. Let’s do it again!

If you’re in the process of defining their go-to-market strategy and want to get a practical feedback. Let’s talk! Grab one of the three spots below.

Each AI Hustle Live session is completely free, uncrecorded and confidential. If you need direction on value proposition, identifying your ICP, experimenting with customer acquisition channels, figuring out retention hacks, making your first hire, or preparing to raise your first round—this session is for you.

Alongside two other founding teams, you’ll receive practical, actionable feedback on how to accelerate your GTM motion and avoid common early-stage pitfalls.

*Spots are first come, first served.

Grab your spot

Analysis

Pricing in Flow.

It looks like the flat-rate software pricing era is over, the latest data shows.

The analysis of 521 SaaS tools across 15 categories found usage-based pricing is now the most common pricing model, used by 31.5% of tools, ahead of per-seat pricing at 29.4%.

The shift is strongest in categories where usage maps directly to cost or value. And so, the usage-based pricing dominates email marketing (92%), analytics (72%), infrastructure (62%), developer tools (52%), and AI/ML (50%) - that’s actually where modern software growth is concentrated as well.

What’s also interesting is that only 26.7% of tools still offer a free tier, while usage limits, metering, consumption thresholds are becoming more common.

And so, a company may still charge per user, but that user fee is no longer the full price. It is just the base layer. On top of it, vendors now add usage charges, credits, quotas, API calls, token consumption, storage, compute, messages sent, workflows run, or overage fees, you name it.

Hiring

This week hiring:

doinstruct ➯ Enterprise Account Executive | WeeFin ➯ Senior Sales Executive | Relay ➯ Chief of Staff to Ops and Engineering.

List your job

Events

This week

Tuesday, June 16, London ➯ GenAI London Warm Up, ➯ Building and Shipping in the AI Era, Paris ➯ Where to build in 2026? | Wednesday, June 17, London ➯ My AI-Native Organisation, Amsterdam ➯ No Such Builders Bootcamp, Warsaw ➯ Agent Builders Dinner | Thursday, June 18, London ➯ ClawBio Hackathon, ➯ Demo Night: Adaptive Software | Sunday, June 21, Berlin ➯ Solopreneur Picnic Meetup.

List your event

Enjoy your Sunday. See you next week!

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

Spanish Cala AI, co-founded by Elisenda Bou-Balust raises a €7M pre-seed round led by Lightspeed Venture Partners; London-based medtech startup Zonova, co-founded by Georgia Fleet and Carla Southworth, secures a £2.1M round led by THENA Capital; UK-based Geordie, co-founded by Hanah-Marie Darley, lands a $30M round led by Balderton for its AI security and governance platform; Also UK-based Atheni, founded by Mackenzie Howe and Louise Ballard, fetches a £350K round from angels for its AI adoption platform; Also UK-based biotechnology startup Mykor, co-founded by Valentina Dipietro and Olivia Page, lands a £4M round led by Clean Growth Fund; And.. also UK-based Voxmind, co-founded by Saritha Kinkiri, picks up a £546K pre-seed round led by Ascension Ventures.

Spotlight

Down the Stack.

Anthropic became the world’s most valuable AI startup this week, nearing valuation approaching $1tn and overtaking OpenAI ahead of its IPO later this year.

For most of the past decade, venture capital was built around the economics of abundance.

Capital was abundant. Cloud computing was abundant. Global supply chains were assumed to be abundant. Talent could be hired across borders. Customers could be reached through digital channels at scale. Software margins were the benchmark, and the most attractive companies were those that could grow without touching too much of the physical world. That’s now changed.

Only three weeks ago, Anthropic signed a landmark computing capacity lease with Spacex, paying around $15 billion a year for access its massive GPU capacity. The reality is also that SpaceX might not be really building infrastructure for its own AI ambitions but turning compute into a business line to fund its Mars project.

What looks at first like a simple compute lease, in reality is something much larger.

AI is no longer just a race between model companies. It is becoming a race between balance sheets, power contracts, data-center footprints and whoever can assemble chips, energy and cooling fast enough.

Just yesterday, SoftBank announced plans for a massive AI data-center project in France. In the past few weeks, Lambda closed a $1 billion to expand its AI factory footprint. CoreWeave expanded its long-term agreement with Meta to roughly $21 billion and French Mistral AI acquired Koyeb.

.. the bottleneck is moving down the stack.

In the old cloud model, a few hyperscalers (think: Google, Microsoft, Amazon) were the landlords but now AI is creating a more fragmented version of that market. CoreWeave, Lambda, Crusoe, Nebius, Nscale are not simply data-center owners. They are all specialised AI infrastructure companies built around the fact that frontier AI demand is growing faster than hyperscalers can satisfy it alone.

This also explains why the economics look so extreme. A $15 billion annual compute commitment sounds extraordinary, but for a leading AI lab, lack of capacity can be more expensive than the capacity itself. If rate limits constrain product usage (especially when you’re prepping for an IPO), compute becomes the growth ceiling.

And so, the AI market is starting to look less like software and more like energy, telecoms and heavy industry. - What used to be the invisible layer underneath software is now becoming the story itself.

AI Hustle

On Advisors.

I spoke with a founder team this week who were debating whether to bring on a very impressive advisor. On paper, the person looked ideal. Senior former operator. Strong logos. Board experience. Good network. Five years ago, it would have been an obvious yes.

But what the team really need right now is not someone to sit on a quarterly call, tell them to tighten the ICP, hire a VP Sales, or introduce three obvious funds. They need someone who could help them make better decisions in a market that no longer behaves like the one those old playbooks were built for.

The classic NED-style advisor who built in the 1.0 era, scaled a seat-based, sales-led company in a zero-rate environment when the market was slow, made a few fund introductions, and asked some board-level questions used to be enough. Not anymore!

They may still have useful scars, but the default playbook is no longer relevant. Today, the GTM expansion is not to optimise the system. It’s to find it.

If you are considering whether you need an impressive advisor for GTM, the answer most likely is: you don’t. Leverage AI to get your structural answers, and pressure-test them (Don’t rely on what ChatGPT tells you, wrestle with it and challenge it!). Then complement it with a sparring partner: someone who understands your industry well from a buyer perspective, but also plays an active role in it.

Only consider people who know how buyers behave NOW, how capital moves NOW, how AI is changing workflows NOW, and where the market is still misunderstood.

Community

Scaleup Soirée — Staying Competitive in the AI Era.

On 16 June, we’re hosting another Scaleup Soirée in collaboration with Rathbones, bringing together fifteen Series A+ founders for an intimate conversation on what it takes to compete in the AI era.

If you are a founder of a Series A+ founder, we have the very last spots left, do join us!

Get your spot

Analysis

Concentrated. How much?

There’s been a lot of talk about market concentration in the past few months. But just how ‘bad’ is the situation? I’ve dived deeper into the data.

In 2025, 70% of US venture funding (c. $200 billion) went to just 389 companies raising rounds of $100 million or more. Within that group, mere six companies (!) captured $90 billion through rounds above $5 billion (yep, Anthropic, OpenAI & co).

Compare it with around 6,000 companies that shared the remaining 30% of funding, (c. $90 billion), across rounds from $1 million to under $100 million, and the picture is clearer. 2025 was one of the most concentrated venture markets on record.

And 2026 is already pushing the pattern further!

Through April, US venture funding was already on par with full-year 2025 levels (yep!), and 80% of startup investment had gone to rounds of $500 million or more across just 29 companies, according to the latest data.

In April alone, AI companies captured 73% of U.S. venture capital, while late-stage rounds took 76.4% of all capital across only only 50 deals.

So far, 2026 looks less like a broad venture rebound and more like an extreme version of the 2025 trend. Funding is rising, but the increase is being pulled upward by a small number of very large companies, mostly in AI and infrastructure-heavy sectors.

Hiring

This week hiring:

Flowww ➯ Sales Development Representative | ingarden ➯ Influencer Marketing Jr Manager | Toloka ➯ Program Director: AI Delivery.

List your job

Events

This week

Tuesday, June 02, London ➯ FR8 x Sequoia: London Tech Meetup, Lisbon ➯ Women in Tech & Cloudflare, Amsterdam ➯ Consumer tech lookout | Wednesday, June 03, London ➯ BGS Best Of Britain Finale, Berlin ➯ Building in the open | Thursday, June 04, London ➯ Design Engineering Night #8 | Friday, June 05, London ➯ NVIDIA Hack for Impact - London.

List your event

Enjoy your Sunday. See you next week!

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

French Pivot, co-founded by Estelle Giuly, lands a €34.4M round co-led by Forestay Capital and Notion Capital, to power enterprise procurement with agentic AI; Italian startup Lexroom, co-founded by Martina Domenicali, bags a €42.9M round led by Left Lane Capital for its AI legaltech solution; Also legaltech German startup LawX, co-founded by Sara Brinkmann pockets a €7.5M round led by Motive Partners to build an AI-powered operating system for notaries.

Greek Skybound, co-founded by Thaleia Misailidou, launches with a $38M fund to back early-stage deeptech founders.

Spotlight

The Stardust.

SpaceX filed its public IPO prospectus on Wednesday this week, asking Wall Street to believe in the biggest story in technology, reportedly seeking a valuation of around $1.5 trillion, a number that would instantly place it beside the biggest names in technology. Nvidia. Apple. Microsoft. Alphabet. Amazon.

From the leaked docs, however, SpaceX does not look like those companies did when they went public.

At the point of listing, Nvidia, Apple, Microsoft, Alphabet and Amazon all had one thing in common: operating leverage. They had scalable models. Most had either profits or small, manageable losses. SpaceX, by contrast, is losing money at industrial scale. In the first quarter of 2026 alone, it posted a $4.28 billion net loss on $4.69 billion of revenue.

That is not the financial shape of a classic tech company. That is the shape of a capital-intensive empire still very much deep under construction (!).

Calling SpaceX a tech IPO gives investors a familiar mental model. Tech companies are supposed to scale quickly. Margins are supposed to expand. Losses are supposed to turn into profits once the infrastructure is built and the customer base is large enough. That was the Amazon argument. It was the Uber argument. It is now the AI argument.

So why should we care, and why is this different?

SpaceX is in reality part rocket manufacturer, part launch provider, part satellite operator, part telecom network, part defense contractor, part national infrastructure asset and part moonshot laboratory. As AI moves from abundance to scarcity — I might write about this next week — SpaceX’s part-infrastructure, part-tech, part-visionary narrative might fit the current investor appetite perfectly!

For the past decade, public investors rewarded asset-light platforms. The best companies were supposed to have software margins, network effects and low marginal costs. - This week, Anthropic was reported to be on track for its first profitable quarter. SpaceX points in the opposite direction. It is deeply physical. It is expensive. It depends on hardware, manufacturing, spectrum, regulation, geopolitics and government demand.

And yet it wants a tech multiple.

And so, this IPO matters beyond SpaceX itself. It may be a test of a new market category: the strategic infrastructure.

If SpaceX can pull it off, it will reset expectations for every capital-intensive technology company waiting behind it. AI infrastructure companies will point to it. Defense-tech companies will point to it. Nuclear startups, robotics companies, satellite networks and frontier hardware businesses will all argue that losses should be judged differently when the market opportunity is strategic enough.

AI Hustle

Dynamic Markets.

I spoke with a founder this week, prepping for her pre-seed round, she did all the ‘right’ things: she knew the market well—category, buyers, competitors, growth rate. But in AI, that is only the starting point (!). The real question is understanding how the market changes when a previously expensive or niche behaviour becomes commercially viable.

I still meet many founders these days that under-read the opportunity. They take current spend, apply a growth rate, and build their commercial case around capturing part of that pool. But AI can create entirely new budget categories, new buyer expectations, new workflows, and new ancillary markets around implementation, compliance, monitoring, verification, orchestration, data quality, and human review.

And so, when you take that account, a market might looks small today only because the behaviour behind it is constrained, too expensive, too manual, or perhaps too dependent on scarce expertise. The question then becomes not ‘how much of today’s market can we capture?’ but ‘what new spend will exist because this capability exists?’

Remember, AI you can absolutely change the underlying unit economics. What was not worth doing at all can suddenly become a recurring workflow and open up a totally new market.

So instead of thinking about the opportunity following Gartner reports, really understand the forces underneath your market. What cost is collapsing with AI? What new behaviour becomes economically rational? What risk is introduced? What layer becomes necessary for adoption? What existing budget gets reallocated and what new budget created?

Remember, as an AI founder, you are not simply entering a fixed market. You are making a bet on how the market will evolve, and then shaping that evolution through product, distribution, and customer traction.

THIS is the commercial story you want to tell.

Community

Picking Winners Amid AI Fever.

I've been covering the competitive nature of startups here a lot, and I'm looking forward to discussing it in depth on Tuesday in London — joining journalist legend Mike Butcher, Director of AI Acceleration Victoria Albrecht at Cleary Gottlieb, and the Founder of AISA, Ozan Dagdeviren.

Join me by registering below.

📅 Tuesday, 26th May 🕕 18:00 - 21:00 📍 The Screening Room, Covent Garden Hotel

Get your ticket

Analysis

The Earnout Retreat.

It looks like earnouts are beginning to ease in startup M&A as venture markets continue to recover from the post-2021 downturn, according to new data.

Earnout provisions showed up in 22% of company acquisitions in 2025, down from 26% in 2024 and 31% in 2023, when financing conditions were toughest. The shift hints that buyers are growing more comfortable paying upfront rather than tying big chunks of a deal to future performance targets.

And the trend is clearest in VC-backed tech acquisitions, where 19% of deals in 2025 included earnouts, compared with 23% in 2024 and 25% in 2023. Contingent structures are still common, but buyers clearly lean on them less than they did at the height of the market correction.

..AI startups are a different story, though. Many acquirers are wary of fully pricing in long-term growth assumptions when revenue durability and monetisation are still uncertain. So for AI startups specifically, earnouts continue to be a fixture, bucking the broader decline.

Hiring

This week hiring:

Open Payments ➯ Head of Sales | ingarden ➯ Influencer Marketing Jr Manager | Toloka ➯ Program Director: AI Delivery.

List your job

Events

This week

Wednesday, May 27, London ➯ AI Showcase: Public Sector Edition, Berlin ➯ Build with Agents — Berlin night | Thursday, May 28, London ➯ Cursor AdTech London Hackathon, Amsterdam ➯ baby vc x Amsterdam | Saturday, May 31, Milan ➯ Codex Community Meetup.

List your event

Have a great rest of the weekend! See you next Sunday.

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

London’s Dessn, co-founded by Gabriella Hachem, raises a €5M round led by Connect Ventures for its production-focused design tool; Also London-based Fifth Dimension, co-founded by Kate Jarvis, raises a €22M round led by HV Capital to scale AI-powered decision intelligence for real assets; German NanoStruct, co-founded by Henriette Maaß, raises a €2.6M round co-led by High-Tech Gründerfonds (HTGF), Bayern Kapital, and the AUXXO Female Catalyst Fund, to detect harmful bacteria in food within hours.

Spotlight

Liquid Gold.

Greg Brockman has officially taken control of OpenAI’s product strategy this week in the latest staffing shake-up, amid the company’s plans to consolidate its product range. The news comes on top of recent reports of Meta confirming a major workforce reshuffle affecting 8,000 jobs, alongside reports of Meta, Microsoft, Amazon, and Alphabet embarking on a combined $700 billion AI infrastructure spending spree.

It’s not bad news for everyone, though — there are clear winners emerging from the headlines too.

This week, it was revealed that in October, more than 600 current and former OpenAI employees sold roughly $6.6 billion worth of stock through a tender offer, averaging around $11 million per employee (!). No IPO. No direct listing. Just enough demand, valuation gravity, and investor competition to create an internal liquidity comparable to a mid-sized sovereign economy.

If that doesn’t raise your eyebrows already, for context: it took Amazon 27 years after founding to mint first employee millioners through public markets. Both Meta and Google needed an IPO.

The old liquidity model has up to this point been relatively linear: raise venture capital, grow aggressively for 6+ years, go public, make first employees millionaires. But what OpenAI is doing — privately, at hyperspeed — is something Silicon Valley has never seen before, at least not at this scale.

And so, we are witnessing first hand a major structural shift in how wealth is being created in technology.

Today’s fastest-growing startups are staying private longer, raising capital at enormous scale, and using secondary transactions to create liquidity internally and privately while continuing to compound value.

OpenAI’s valuation discussions last week pushed the company toward the $950 billion valuation — that’s 6x value of Salesforce, 8x of Intel, and of many public global banks, despite (!) still being private.

At that scale, employee compensation is starting to function more like an internal capital market. And when you add recent data showing that AI startups are now capturing 61% of global venture capital — despite representing a far smaller percentage of total startup volume — you can see that we are witnessing something structural.

For context, private company secondary transactions crossed roughly $240 billion globally last year, with Stripe, Databricks, SpaceX, Anthropic, and OpenAI all increasingly relying on structured tender offers instead of public listings.

And what’s making the wheel spin even faster is the infrastructure buildout directly affecting compensation and liquidity dynamics.

Top AI engineers now routinely receive annual compensation packages exceeding $5–10 million across equity and cash (see: Week 116).

And so, in a few years’ time, OpenAI’s $6.6 billion employee tender may be remembered less as an isolated event and more as a preview of what the next generation of technology wealth creation looks like: private, concentrated, and dominated by a small cluster of AI players absorbing extraordinary amounts of global capital, compute infrastructure, engineering talent, and labor-market pricing power .. all at once.

AI Hustle

The Productivity Dilemma.

I spoke with three founders this week, individually quoting consistent feeling of ‘swimming upstream’, being ‘overwhelmed’, feeling ‘exhausted’. A coincidence? Weren’t we promised that AI would make our lives easier?

For years, we imagined productivity gains would lead to more free time. But see, technology rarely works that way (and especially in competitive industries).

Every efficiency breakthrough simply resets expectations upward.

Think about it: email didn’t reduce communication. It multiplied it. Slack didn’t create calm. It accelerated response culture.

Now AI is compressing cognitive work itself.

Coding, research, marketing, recruiting, product iteration — entire workflows are collapsing from days into hours and being absorbed immediately by the market.

If one founder can suddenly do the work of ten people, investors expect faster growth. If engineers ship faster, roadmaps expand. If content becomes easier to produce, your customers expect more of it.

This is why AI may not initially feel like an ‘energy-saving’ revolution but more like permanent acceleration.

The real AI value is not in automation but the removal of organisational friction. Smaller teams can now operate at the scale of much larger companies, which is already changing hiring, compensation, and expectations across tech.

So what should you do while flying through this storm?

Firstly, don’t let the wind sweep your wings!

Remember, markets always optimise for output. Technology never optimises for rest.

And so, when every founder gains access to the same AI tools, the real differentiator is no longer pure productivity — it’s knowing what deserves your attention in the first place. Your judgment and clarity of thinking matter more now than ever.

Protecting your energy and sharpening your focus can now be a real strategic advantage.

Community

Google For Startups x Energy Startups

Hey founders! Our friends at Google have just opened applications to
the Google for Startups Accelerator: Europe & Israel.

Who should apply?

• Energy efficiency and utilization: startups developing AI solutions that reduce energy burdens for households or lower costs for industries.

• Grid modernisation and enhancement: startups building AI solutions and Grid Enhancing Technologies.

• Demand flexibility and optimization: startups.
  all leveraging Cloud or AI/ML in their product, service, or operations

📍 Next info session on May 20th [2 pm CET]

Apply Now

Analysis

On Defence.

It looks like one of the most historically defamed categories in venture capital — defencetech — is rapidly becoming one of Europe’s hottest investment areas.

According to new data, European defense, security, and resilience startups raised a record $8.7 billion in 2025, up 55% YoY and nearly 4x higher than five years ago. Defense tech now represents roughly 13% of all European VC funding and 43% of deeptech investment (!).

The biggest funding flows are now going into battlefield AI, autonomous drones, ISR systems, satellite intelligence, electronic warfare, and dual-use robotics. Investors increasingly view the category less as traditional defense and more as AI infrastructure tied to sovereignty, resilience, and geopolitical independence. And capital is concentrating aggressively into a handful of winners. The five companies: Helsing, Quantum Systems, Iceye, Destinus, and Tekever account for almost 60% of all European defense-tech funding since 2021.

Geographically, Germany is emerging as a leader, capturing roughly 90% of Europe’s defense-tech capital last year, driven largely by Munich-based AI and autonomous systems startups.

.. and the macro trends are only accelerating the trend. NATO-backed funds now have more than €1 billion dedicated to defense-focused deeptech, as European governments continue ramping military and security spending targets higher and higher.

Three years ago, many European VCs wouldn’t touch defense. Today, it’s becoming one of the continent’s fastest-scaling VC categories.

Hiring

This week hiring:

Open Payments ➯ Head of Sales | ingarden ➯ Influencer Marketing Jr Manager | Toloka ➯ Program Director: AI Delivery.

List your job

Events

This week

Monday, May 18, London ➯ Money20/20 Europe: Pre-Event Mixer, Paris ➯ Build Your GTM Agent | Tuesday, May 19, London ➯ AI Agents 101: Become an Agent-First Founder, Amesterdam ➯ AI Meets the Brain | Wednesday, May 20, Lisbon ➯ AI Builders Lisbon | Thursday, May 21, London ➯ Founders Circle: Founders Truth, Berlin ➯ Women in Deep Tech Berlin.

List your event

Have a beautiful Sunday. See you next weekend.

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

Germany-based medtech startup Sedivention, co-founded by Ute Nollert, picks up a €2.9M round led by bmp Ventures; UK-based Nyobolt, co-founded by Clare Grey, lands $60M round led by Symbotic; Spain’s Humara, co-founded by Laura Rodríguez Álvarez, fetches €1.2M led by Impact Shakers, to cut waste plant design cycles; Swiss Moonlight AI, co-founded by Nicole H. Romano, raises €2.8M co-led by Lotus One Investment and VP Venture Partners to turn routine blood and cytology imaging; French OpsMill, co-founded by Karen Gallantry, lands a $14M round led by IRIS for its enterprise infrastructure data solution; Finland-based deeptech startup Reduciner, co-founded by Johanna Grönroos, secures a €3.6M round from Voima Ventures and Lifeline Ventures to commercialise its carbon conversion technology; German smart city startup ENVIOTECH, co-founded by Linh Pham, fetches €1M led by Jürgen Fitschen for its intelligent retrofit solutions for public street lighting.

Spotlight

The Race of Spades.

Anthropic launched a PE-backed vehicle supported by giants such as Blackstone, Hellman & Friedman, Goldman Sachs, Apollo, GIC, and Sequoia this week to gain access to the hundreds of mid-market companies sitting inside PE portfolios. OpenAI is now reportedly building a similar structure with TPG, Brookfield, Bain Capital, and others, making the race between OpenAI and Anthropic even more interesting, and underwriting the notion that the big AI model race is no longer just about building the best models (see: Week 184) but increasingly about who can scale enterprise adoption fastest before the IPO window opens.

Why does it also matter?

For much of the AI boom, PE firms have been somewhat struggling to find its footing, amid a record backlog of roughly 32,000 unsold portfolio companies worth approx. $3.8 trillion. Venture capital dominated the infrastructure layer, hyperscalers controlled compute, and the biggest value creation stories centered around foundation models and startups.

Now, it looks like doing what PE does best: operational scale and distribution is about to turn the tide for the industry.

For context, PE firms now manage roughly $15 trillion in assets and oversee roughly 30,000 portfolio companies globally — a massive distribution network of businesses scrambling to adopt AI without the internal engineering talent to do it alone. For OpenAI and Anthropic, that’s a lucrative shortcut into enterprise adoption at massive scale!

With their IPOs fast approaching, one thing to remember is that public markets won’t reward AI companies based purely on technical leadership forever. Good old recurring revenue and profitability will eventually matter most.

And so, if AI tools can improve margins, automate workflows, or unlock new revenue streams across PE portfolio companies, the impact flows directly into PE higher exit valuations - that’s a very strong alignment.

And, over time, the opportunity could become even bigger: for OpenAI and Anthropic, the services and workflow layer is where recurring revenue, customer dependency, and long-term enterprise relationships tend to live and that’s how the big bucks are found.

In many ways, this resembles early enterprise software platform wars. The winners were rarely the companies with the best technology (think: Lotus 1-2-3 vs Microsoft). They were the ones that become embedded into how businesses actually operate.

AI Hustle

The Product Mind.

One of the frequent traps I see AI-native founders fall into these days is mistaking speed for a product.

Today, anyone can build fast. AI copilots, APIs, and low-code tools make it possible to ship features in hours instead of weeks. You would be surprised how many founders I speak to these days who boast about how many features they built in the past ‘couple of weeks’. - I used to be impressed.

But see, most startups do not fail because they move too slowly. They now fail because they build too much of the wrong thing and in the wrong direction.

I see many founders these days blur the line between a feature and a product. And that’s a major problem. A feature is something users can technically use. A product is something they would genuinely miss if it disappeared.

Don’t become yet another workflow enhancement founder! .. How do you know whether you already are? 👀

First, look at urgency. If your product disappeared tomorrow, would users actively look for an alternative, or simply move on?

Second, look at workflow gravity. Features sit on the edge of someone’s workflow. Products sit right in the centre of it. If users occasionally test your tool, you probably have a feature. If it becomes part of how work consistently gets done, you are building a product.

Third, pay attention to compounding value. Features create isolated moments of usefulness. Products become more valuable over time through habits, integrations, context, collaboration, or accumulated data.

- This distinction matters because AI capabilities are commoditising incredibly quickly. Remember, the moat rarely comes from the model itself. It comes from distribution, customer trust, workflow integration, and deep understanding of user behaviour.

Also,

the best founders I meet these days focus on the smallest possible surface area that creates disproportionate value. Instead of asking, ‘What else can we do?’ they ask, ‘What is the one thing users cannot live without’.

Those founders spend less time obsessing over what they can build and more time obsessing over what customers actually need.

They study friction points carefully. And most importantly, they resist the urge to build before they assess the urgency and strategic customer value.

Community

Coming to America Dinner.

It was fantastic to bring together founders looking to raise capital from American investors and expand to the U.S. at the dinner we co-hosted a couple of weeks ago in London with Wilson Sonsini—a leading Silicon Valley law firm behind top tech companies (think: Apple, Google, LinkedIn, Netflix, Eleven Labs, Tesla, Twitter) and its UK team specialising in US-UK fundraising and expansion—to answer your questions.

Missed the dinner and still have questions? The Wilson Sonsini team has put together a very comprehensive FAQ with up-to-date information to answer them.

Are you looking to expand or fundraise in the U.S. in the coming months and would like an introduction to the Wilson Sonsini team? Get in touch.

Analysis

Say “A”.

You might already be well aware of the fact that Series A is perhaps the hardest round to raise. But it looks like where you’re building matters more than ever.

The latest data paints a pretty clear picture: Series A investors are getting increasingly concentrated around a handful of sectors.

Right now, AI infrastructure is leading the pack. Not just foundation models, but the entire ecosystem around them — think: world models, vision models, developer tooling, inference layers, and other “picks and shovels” startups. Investor appetite here is massive. Semiconductors, hardware and cybersecurity are also huge winners.

Meanwhile, sectors like food & beverage, edtech, and medical devices are seeing far less enthusiasm. Even with the AI boom, edtech valuations and deal volume haven’t meaningfully bounced back.

The bigger shift may simply be that AI is no longer a differentiator. So far in 2026, around 80% of software startups raising were AI companies in some form. AI is no longer the wave — it’s the default.

Which raises the question: does this create an opportunity for the non-AI contrarians?

One thing is clear: the money is still out there, and some companies are raising at wild valuations earlier than ever. But this is still an extremely tough fundraising market investor attention is incredibly fickle.

Hiring

This week hiring:

doinstruct ➯ Senior Customer Success Manager | HelloBetter ➯ Team Lead Sales | Toloka ➯ Freelance Tech Recruiter.

List your job

Events

This week

Tuesday, May 12, London ➯ The Founder Mixer - London VC Week, Netherlands ➯ AI Meets the Brain, Netherlands ➯ Lenny’s Newsletter Meetup, Paris ➯ Build Your GTM Agent | Wednesday, May 13, London ➯ five degrees: Founder Showcase | Thursday, May 14, London ➯ Startup Roadshow - Robotics, Barcelona ➯ An Investor Roundtable | Friday, May 15, London ➯ AI x Weather x Climate Demo Night.

List your event

Enjoy the week ahead! See you next Sunday.

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

French Cleo Labs, co-founded by Naomie Halioua and Anaelle Guez secures €1.5M led by Larry Berger to scale AI-driven product compliance globally; Prague-based payments infrastructure startup Tapaya, co-founded by Laura Ďorďová picks up a pre-seed round led by Passion Capital; Berlin-based Patronus, co-founded by Rebecca Qian, raises a €11M round led by 3TS Capital Partners for its senior-friendly emergency smartwatch and family app; Germany’s mbiomics, co-founded by Dr. Laura Figulla raises a €30M round led by MIG Capital to scale microbiome therapeutics for chronic diseases; German Primogene, co-founded by Linda Karger, raises a €4.1M round led by High-Tech Gründerfonds to scale enzymatic biomanufacturing of complex bioactive molecules; Dutch quantum startup Groove Quantum, (Female Innovation Index 2026) co-founded by Anne-Marije Zwerver, raises a €16M led by Innovation Industries to advance scalable chip manufacturing; UK-based in-space drug manufacturing company BioOrbit, co-founded by Katie King, lands a £9.8M round led by co-led by LocalGlobe and Breega; Also UK-based healthtech startup auryx, co-founded by Kayla-Jade Butkow, Cecilia Mascolo and Erika Bondareva raises $2M round led by Celero Ventures to turn earbuds into health monitors.

Copenhagen-based KOMPAS VC, co-founded by Talia Rafaeli, closes €160M Fund II to back industrial tech startups.

Spotlight

The Hard Grip.

Tim Cook announced last week he was stepping down as Apple’s CEO after 15 years, leaving an astonishing legacy behind him - taking Apple from $350 billion to nearly $4 trillion in market cap, growing profits eightfold (!), and scaling it to over 2.5 billion devices. Cook is handing the job to John Ternus, a 25-year Apple lifer and a hardware expert, sending a clear signal: while everyone else is pouring money into AI infrastructure Apple is doubling down on hardware, betting AI will live inside devices — iPhones, Macs, wearables — not just in the cloud.

Why does it matter?

Because hardware is still only part of the story. Not many people know that in 2025 alone, the App Store made about $31.7 billion for Apple — roughly a third of its services revenue. And so, what might look like just a hardware-first strategy is actually something far more layered: a tightly integrated system where devices drive distribution, and distribution drives recurring revenue.

Now, in the past quarter, App Store submissions have been up 84% — staggering, driven by AI “vibe coding” tools that let users generate apps at scale — think: Replit, Base44, v0.dev. Surely that’s good news for Apple — not so easy.

Vibe-coding apps are actually bad news for Apple primarily because they bypass the strict, curated control Apple exerts over its iOS ecosystem, posing security risks and undermining the App Store’s review process. On paper, it’s policy. In reality, it’s control. If apps can create apps, Apple loses its grip on the ecosystem it has meticulously created for almost two decades!

So this is the real story. Cook built the machine. Ternus now runs it. And Apple is trying to hold control of both the devices and the way software gets made on them.

If Apple manages to keep that control (and that’s a big if), it doesn’t just stay relevant in the AI era — it defines how that era is experienced: tightly integrated, device-first, and on its own terms. If it doesn’t, the center of gravity shifts away from Apple’s walled garden toward a more open, faster-moving layer of software creation — one Apple no longer fully owns.

AI Hustle

The Messy Reality of Demos.

I spoke with a founder this week who wanted my feedback on his demo - all prepared, link was queued up, log in details entered. The demo was polished. No comments on that. What stood out was the lack of reality.

I keep seeing founders present their AI products and the reaction is electric. People lean forward, eyes widen, someone inevitably says “whoa.” And then, a few months later, retention falls apart and founders are lost. Remember, demos are designed for one type of intelligence, while retention depends on something entirely different.

In a demo, everything is controlled. The input is carefully chosen. You know exactly what will happen. The AI completes the task in one clean pass, delivers something impressive, and gets applause. It proves a point: this can work.

But real-world is messy. Users don’t show up with perfect prompts. They’re inconsistent. They break the flow. They ask odd follow-ups. They leave and come back later with no context. At that point, the question is no longer can this work? but does this actually work for me, over and over, in real life?

The AI products that people stick with aren’t flashy in demos. They’re steady. They’re clear about their limits. They don’t try to do everything. They handle most of the task and make it easy for the user to complete the rest. And crucially, they make users feel capable, not replaced.

The ones that impress on stage tend to do the opposite. They overpromise. They act more certain than they should. They hand over a polished result when what the user really needed was help thinking things through.

From experience, if your demo only shines with perfectly crafted input, you’re not looking at a real product yet. It’s a proof of concept dressed up as one.

The disconnect between a great demo and real retention isn’t about what the technology can do. It’s about how it’s designed to be used .. and whether it accommodates the messy scenarios of your users.

Community

Panathēnea x Female Foundry

The original Greek startup festival Panathēnea is bringing together 10,000 founders, investors, creators and operators is taking place on 27–29 May in Athens and we have 20% discount on all ticket types.

To get the 20% discount on the tickets, use the code: EQUISSCOREEP26 and this link to book your tickets:

Get your tickets

Analysis

Spinouts Run High.

It looks like even as total dealmaking slows, European university spinouts are pulling in a record share of funding—€3.1 billion so far this year, accounting for 12% of all VC investment across the continent, an unprecedented high, according to the fresh data.

Europe hosts ~30% of the world’s top deep-tech universities and produces twice as many science and engineering graduates as the US with structural barriers also easing. - In the UK, for example, average university equity stakes dropped from 21.5% in 2023 to 16.1% in 2024—the lowest in a decade.

Hiring

This week hiring:

Miss Moneypenny Technologies ➯ Junior Account Executive | TOLOKA ➯ Freelance Tech Recruiter | Quantexa ➯ Demand Generation Manager.

List your job

Events

This week

Tuesday, May 5, Berlin ➯ Before You Sign: How Term Sheets Actually Work | Wednesday, May 6, Paris ➯ Hackathon PoC | Thursday, April 7, London ➯ Future of Frontend, Stockholm ➯ Scale with Sweden, Lisbon ➯ AI Innovation Day: Real World Automations.

List your event

It’s been a good week. Enjoy your Sunday!

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

German ATMOS, co-founded by Marta Oliveira, lands a €25.7M round led by Balnord and Expansion Ventures to establish Europe’s first orbital return infrastructure from testing to commercial operations; Swiss Bubble Robotics, co-founded by Patricia Apostol, fetches a $5M round led by Episode1 Ventures, Asterion Ventures, and Norrsken Evolve to scale its autonomous ocean robotics platform; University of Bath spinout Naturbeads, co-founded by Giovanna Laudisio and the late Janet Scott, picks up a €4.1M round led by the European Structural Fund to support construction and commercialisation of its biodegradable ingredients technology.

London-based Passion Capital, co-founded by Eileen Burbidge, closes a €46M fund to back AI, Fintech and Enterprise Risk start-ups across the UK and Europe.

Spotlight

In the Middle.

Global startup funding surpassed $300 billion between January and March — a massive 2.5x increase over the $118 billion raised in the previous quarter, and a single-quarter total that exceeded every full year of global VC activity prior to 2019.

But when you look closer, the spike was clearly driven by four behemoth deals — OpenAI ($122 billion), Anthropic ($30 billion), xAI ($20 billion) and Waymo ($16 billion) — that drove 63% of all venture funding in the quarter. Strip them out, and the market starts to look a lot more familiar.

Also this week, Jeff Bezos’ AI lab, Project Prometheus, raised $10 billion at a $38 billion valuation, becoming one of the most valuable five-month-old startups ever funded — and that’s just one of the things Bezos has been busy with. In parallel, Bezos is also reportedly looking to raise a $100 billion (!) fund focused to acquiring manufacturing companies and applying AI to them. .. All of this while it’s also been reported that Amazon’s own AI boom has been creating a mess of duplicate tools and fragmented data internally.

Nevertheless, apart from large scale up rounds, mega-funds are also having a moment.

In just a couple of years, Ardian raised $30B, Thoma Bravo $24B, EQT nearly $25B. $10B+ funds used to be outliers — in 2025, nearly half of all capital went to the top 10 funds.

But raising capital at that scale is only half the equation. Deploying it effectively, without compressing returns, is where things get harder.

SoftBank’s Vision Funds are the obvious case in point. Vision Fund I, launched in 2017 with nearly $100 billion, delivered some standout wins and very clear missteps — including the collapse of WeWork, which alone led to losses exceeding $14 billion. The broader Vision platform posted a $32 billion loss in 2023, and nearly a decade on, Vision Fund I has returned ..just 0.61x invested capital.

Mega-funds promise access to transformative opportunities, but they also demand precision. Large checks, concentrated into fewer companies, increase the reliance on category-defining outcomes — while raising the risk of overpaying and overcapitalisation. When bets don’t land, even a handful of losses can damp overall returns.

Now, add AI to the mix — where capital is flowing at unprecedented speed, but realised returns remain limited — and the result is a growing imbalance between how much is being invested and how much has actually been proven out.

So what looks like a record-breaking quarter is, in reality, a signal of a deeper structural shift. Venture capital isn’t scaling — it’s concentrating and reorganising itself. On one end, a handful of scale-ups absorb unprecedented levels of capital; on the other, early-stage AI startups are pulling in more funding from VCs eager not to miss the next breakout.

What about the middle? Hear the crickets.

As capital clusters around clear winners on one side and bold early bets on the other, the middle is starting to feel eerily empty.

AI Hustle

The Curious Minds.

I spoke with yet another founder this week who’s been seeking hiring advice. He recently raised his first $5M round while in stealth and has been struggling to attract the right talent.

I’m not a recruitment expert, but what I do know from being in the AI building space since 2022 is that AI hiring is a balancing act. You’re basically combining a small number of highly specialised technical hires with a broader team of AI-enabled generalists. You’re continuously reassessing what needs a human, what can be automated, and what sits somewhere in between. And all of that while battling for talent against hyperscalers and scale-ups.

But remember:

On one end, there’s strong demand for deep technical talent — ML engineers, applied AI specialists, data infrastructure leads — especially if you’re building AI at the core of your product. These hires are expensive, highly competitive, and often define your technical moat (more on that in the coming weeks).

But for most startups, the bigger unlock lies in hybrid profiles — people who aren’t necessarily “AI engineers” but know how to work with AI effectively.

And so:

Traditional signals — degrees, even past job titles — matter less today than ability to use AI tools to produce outcomes. Note: Seek portfolios, side projects, and practical demos from your new hires. Instead of theoretical interviews, try real-world tasks: ask them to solve problems using AI tools, design workflows, or show how they would automate part of the business. Insead of “do you know AI?” — but “show me how you use it to create leverage?”

The best AI hires are curious, hands-on, and constantly experimenting.

Lastly, pay attention to overcorrecting. Not every role should be replaced or heavily augmented by AI. The goal isn’t to minimise headcount at all costs — it’s to maximise output per person. The best teams aren’t the smallest, but the most leveraged.

Community

Invitation: Scaleup Soirée — Staying Competitive in the AI Era.

On the 19th of May, in collaboration with Rathbones, I’m hosting an exclusive dinner for fifteen scale-up founders to share their strategies for staying competitive in the AI era.

The evening will be a great opportunity to exchange ideas and make connections in a relaxed and intimate setting.

If you are a founder of a Series A (or later) scale-up, do join us!

Get your spot

Analysis

If not AI.

It looks like you don’t have to build an AI startup to raise venture capital — but in some sectors, it’s quickly becoming the default.

AI startups are clearly dominating parts of the early-stage landscape, but the reality is more uneven than the headlines suggest, the latest data shows, with the pace of adoption varying significantly depending on the sector, especially at the pre-seed to seed stages.

If you’re building in marketing, sales, dev tools — AI isn’t optional, it’s expected. In core SaaS, AI is no longer a differentiator, it’s the baseline. Around 63% of rounds are going to AI startups, and an even larger share of the capital.

A similar dynamic is playing out in edtech, where again more than 60% of early-stage rounds are now going to AI-first companies.

Fintech and hardware are moving in the same direction, just at a slower pace. Roughly 40% of companies raising in these categories are AI-driven — meaningful, but not yet dominant.

And then there are the lagging sectors. Medical devices, energy, biotech — here, fewer than one in four startups raising early capital are classified as AI. Part of that is structural. More physical, regulated, or capital-intensive industries don’t absorb new technologies at the same speed as software. But it also highlights where the next wave of opportunity may be building.

So no — you don’t have to be an AI startup to raise. In some sectors, AI is already table stakes. In others, it’s still early — which may be exactly where the upside is.

Hiring

This week hiring:

Zelt ➯ Sales Development Representative (fast track programme to Account Executive) | LIBEEN Smart Housing ➯ Content Creator & Social Media Manager | Quantexa ➯ Demand Generation Manager.

List your job

Events

This week

Monday, April 27, Amsterdam ➯ Next Frontier AI, Hamburg ➯ Social Impact Republic | Tuesday, March 28, Paris Fundraising vs Bootstrap, London ➯ Building Breakout Companies | Wednesday, April 29, Budapest ➯ Claude Code for Everyone, Helsinki ➯ Femcode Collective x Dottir | Thursday, April 30, Paris ➯ Pitch Roast #1, London ➯ Angel & VC Pitch Night | Friday, May 01, Amsterdam ➯ Creativity in the Intelligence Age.

List your event

Have a good Sunday. Enjoy the week ahead!

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

♡ Thank you Hannah for the research!

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.

Female Foundry - the Future of Venture is Here.

In the news

Stockholm-based Enkei, co-founded by Lovisa Sunnerholm and Miriam Bichsel, picks up a €3M round led by RadCap to advance its circular material production, startup; Finland’s Audicin, co-founded by Laura Avonius and Dr. Victoria Williamson, secures a $1.9M round led by private investment with backing from Petteri Lahtela and Virpi Tuomivaara to help commercialise their sleep headband using nervous system regulating neurotechnology.

Bootstrap4F, co-founded by Fatou Diagne and Stéphanie Heller, announces first close of ‘Women Backing Women’ fund at £130m of UK’s flagship fund of funds to back female investors and founders.

Spotlight

The Turning Point — The Turing Moment.

SpaceX filed for its IPO on Wednesday, the same day the U.S. sent astronauts around the Moon for the first time in more than 50 years — a ‘coincidence’ that says a lot about where space exploration is heading. Estimates suggest that Spacex will aim to raise $75 billion at a valuation of $1.75 trillion, making it the largest IPO in history and catapulting the rocket company to one of the highest market caps on a stock exchange.

Why does it matter?

This could well be the last time we saw NASA go to deep space without major help from VC-backed companies.

NASA’s return to the Moon with Artemis II has been decades in the making (and is expected to cost over $90 billion this decade). The mission traces back to the early 2000s, when the agency began developing the Space Launch System (SLS) and Orion spacecraft. By 2010, the program was already well over budget and had to be scaled back — which meant that for the first time ever, NASA began supporting private rocket companies, a decision that helped keep SpaceX alive and opened the door to a new wave of VC-backed space startups. Since then, tens of billions of dollars have flowed into the sector.

That split — between legacy and new — is still visible today. Even though SLS, built by Boeing and Lockheed Martin, is today the most powerful operational rocket in the world, it represents a traditional way of space exploration thinking: technically impressive, but slow, expensive, and difficult to iterate on. SpaceX, by comparison, has taken a totally different approach — rapidly iterating, focusing on reusable rockets, and driving launch costs down by as much as 90%. Today, SpaceX accounts for over 70% of all orbital launches.

One of Musk’s next ambitions for SpaceX — recently merged with xAI — is to build data centres in orbit to support the growing demands of AI. This week, space data centre startup Starcloud raised $170 million at a $1.1 billion valuation — a signal that investors are starting to take this new category seriously.

In the meantime, space is no longer just about exploration ..again. China is also aiming to land astronauts on the Moon by 2030 — the same timeline NASA has been working towards.

And so, as SpaceX has already reshaped the economics of getting to orbit, what comes next — returning humans to the Moon — will be a much bigger test.

If it works, it won’t just mark a new era of space exploration, but also a turning point in how we build: proving that the VC-backed model can extend far beyond software, into solving some of the most complex challenges of our times.

AI Hustle

The other side of the Moon.

The world stood still on Wednesday at 11:05 PM BST as Artemis II lit up the sky, marking a historic mission to explore the far side of the Moon for the first time in over fifty years.

For those who know me, also know that I’ve always been fascinated by space. My grandad was an astrophysicist, and from a young age I learned about planets, black holes, and the vastness of what lies beyond us.

As I’m writing this, the astronauts are about 30 hours away from seeing the other side of the Moon up close, but what I find most inspiring isn’t just the mission itself, but the four people on board. (get to know them here: 3hrs:36mins:15sec.). Do they have doubts? Of course. Do they know it’s dangerous? Absolutely. Do they still believe it’s worth doing? Without doubts.

And that’s what building in the age of AI feels like right now.

You are operating in a moment of extreme acceleration. The pace of change is unlike anything we’ve seen before. Entire industries are being reshaped in months and no one has full visibility on what the next three, four years will look like.

How do you build in the times like these? The answer isn’t to have perfect clarity. The astronauts don’t. - It’s not to eliminate risk, you can’t. It’s to believe that what you’re building is worth the journey — even without full visibility.

The founders who will define this era won’t be the ones who waited for things to settle. They’ll be the ones who focused less on hype and more on real problems. The ones who understood that tools change, but critical thinking compounds. The ones who stepped into the uncertainty, asked better questions, and kept going despite not having all the answers.

Analysis

Fast and High.

It looks like global venture funding has kicked off the year at record levels, with investors minting early-stage unicorns at a remarkable pace, according to the newly-released data.

47 seed- and early-stage companies hit $1 billion valuations in Q1 alone, putting 2026 on track for the largest cohort of young unicorns ever recorded.

This follows a strong 2025, when 59 companies reached unicorn status — up about 50% from 2024. While numbers have fluctuated over the past decade, one thing is clear: nearly all recent early-stage unicorns are AI-focused (think: Project Prometheus (launched by Jeff Bezos in November) and London-based Nscale, which has raised over $5 billion three weeks ago).

This momentum isn’t surprising — 80% of global venture funding last quarter went into AI. With record funding and rapid momentum, it does raise the question of whether we’re nearing a peak in AI-driven valuations? What do you think?

Community

Female Foundry x Tech.eu Summit London 2026

Female Foundry is a Community Partner of Tech.eu Summit London 2026, taking place on 21-22 April at the QEII Centre, London that will bring together leading founders, investors and technology professionals from across Europe and beyond for two days of in-depth discussions, practical insights and meaningful connections, and we have 3x FREE tickets to give away to our community.

Book your ticket

To enter the FREE ticket draw, submit your interest below. Deadline: April 10, 10pm BST. T&C of Tech.EU Summit apply. I will announce the winners here in two weeks and directly via email. Good luck!

Enter the Draw

Hiring

This week hiring:

Zelt ➯ Product Associate (Entry Level) | Cambridge GaN Devices Ltd ➯ Digital Marketing Specialist / Manager | snuggs ➯ Influencer Marketing & Partnerships Specialist.

List your job

Events

This week

Monday, March 06, Paris ➯ Prime Intellect | Tuesday, March 07, Stockholm ➯ Deep-Tech and Venture Capital, London ➯ Deep Tech, Scale, and Europe's Ambition | Wednesday, April 08, Amsterdam ➯ Founders Session: GTM by Design, Helsinki ➯ Frontier Interfaces Workshop | Thursday, April 09, London ➯ What Makes a Black Swan?, Helsinki ➯ Europe's next unicorn is hiring | Friday, April 10, Zurich ➯ Ethereum Zuri 2026.

List your event

I’m off for two weeks, enjoy your Easter weekend!

Agata

Written by Agata Leliwa Nowicka, a startup adviser, a two-time tech entrepreneur, and the founder of Female Foundry based in London. Agata holds an MBA from INSEAD and Wharton Business School.

Would like to get a 1-1 feedback on your fundraising or business strategy? Book a 1-1 call with me here.

Suggestions? Drop me an email.

♡ Thank you Hannah for the research!

Check femalefoundry.co for more fundraising tools and investor content. View other Female Foundry articles.