I pulled the data from Eurostat and came up with the following:

// observations 1

Taken together, these 21 economies generate apx. €3.1 trillion in GDP.

Poland is the clear outlier. Its economy is 2.4 times the size of Romania’s (the region’s second-largest), and accounts for roughly 30% of CEE’s combined GDP. Poland alone is almost as large as Romania, Czechia and Greece combined.

Poland being the largest was not a surprise to me, but the size gap was.

Behind it comes Romania at €380 billion, Czechia at €347 billion and Greece at €249 billion. Czechia stands out: its economy is almost as large as Romania’s, despite having little more than half of its population.

Hence, I started looking into the GDP per capita:

// observations 2

There are a few findings:

• I was surprised to have Cyprus on top of the list;

• It’s clear that countries outside of EU are lagging behind (all are below Bulgaria);

• Romania and Bulgaria rank at the bottom within the CEE EU members;

• The largest economies - Poland, Romania and Greece are the largest due to their size. They don’t outperform per capita. Czech Republic is the only one that stands out. It’s both 3rd largest economy and 3rd highest nominal GDP per capita;

• Ukraine ranks last. It was already low before the full-scale invasion, but the war makes its current position difficult to interpret.

// what’s behind these economies?

CEE’s largest economies are driven primarily by industry and by trade, transport and hospitality, with real estate also playing an important role in some countries. Each has its own particular mix. Just as an example Czechia has the biggest industrial component in GDP, while Greece is especially exposed to real estate and tourism-related activities.

But this is a topic to dig into some other time!

Data methodology: figures represent nominal GDP at current prices for 2025. The primary source is Eurostat, however IMF was used for Ukraine (due to gaps in Eurostat).

I always knew Romania and CEE did well in math olympiad. But I digged a bit deeper to see the actual data. Here’s what I found:

// observations

Romania ranks first in the region (😁), with 156 medals won between 2000 and 2026. Interestingly, the differences among the top six countries are relatively small. Hungary and Ukraine are tied for second place, followed by Bulgaria, Belarus and Poland.

Adjusted for population, Estonia comes first, followed by Moldova, Croatia, North Macedonia and Bosnia and Herzegovina. This comparison isn’t that accurate, as every country can send the same maximum of six students (so you don’t really scale with population), but it still highlights how some of the small countries perform pretty well.

Another thing that stood out to me is how well all of Romania’s neighbors perform. Hungary and Ukraine are tied for second place, Bulgaria ranks fourth, Serbia tenth and Moldova twelfth. There seems to be a math cluster around Romania! Maybe also because the first International Mathematical Olympiad was organized in Brasov, Romania, in 1959, with seven participating countries - almost all from Eastern Europe.

// why 2000 onwards?

First, it gives us a clean look at performance in the 21st century. It also avoids some of the historical complications created by countries previously competing as part of Czechoslovakia, the Soviet Union or Yugoslavia.

One caveat remains: Serbia and Montenegro competed together until 2005, winning 26 medals during 2000-2005. Serbia has nevertheless won another 102 medals since competing independently from 2006, placing it tenth despite having six fewer editions.

// global rankings

I was also curious to see how these results compared globally.

Here is the global top 15:

The result is quite impressive: Romania (6th!), Ukraine and Hungary are all in the global top 10, while Bulgaria brings the region to four countries in the top 15.

If we extend the comparison to include Turkey, five of the global top 15 countries come from CEE and Turkey. That is exactly one-third of the ranking.

// conclusion

The broader conclusion is difficult to miss: CEE represents one of the deepest and most consistent clusters of mathematical talent in the world.

And, yes, I am happy that Romania came first.

Would you invest?

// the story

Let’s add some context. Here is how the company developed during its first ten years:

In its 10th year, the business had reached apx. €25 million in revenue and was generating north of €4 million in annual net profits. Employing north of 1000 people.

I’d argue that’s a good outcome. The company was fully bootstrapped and only operated locally (Romania). Not a venture-scale opportunity, but it was certainly the type of business many people would be happy to own.

For a small private equity fund, it could have been an interesting take;
For an angel investor, a pretty good bet;
For the founder, a very strong result - building a company that generates almost €4 million in annual profit within ten years is no small achievement - particularly without raising outside capital.

However, an operational headache, probably!

But this is what happened next:

Between 2020 and 2025, the company grew its revenue more than sixfold, from approximately €25 million to €160 million. Annual net profit increased to apx. €10 million. In 2025 alone, revenue grew by almost 30%.

So, what does the company actually do?

It sells pretzels. More specifically, it sells the Romanian covrig, sold from small, fast-service locations across the country. See it here:

The company is called Luca.

On the surface, it may sound like a super boring business. The product has existed for generations. There is no technological breakthrough, no new consumer category and no complex intellectual property.

But the founders took a large, fragmented market and turned it into a highly scalable retail machine: they standardized the product, built a recognizable brand people love and created a format that could be replicated hundreds of times. And they sold an affordable product that customers buy “recurringly”.

More recently, Luca has also started expanding internationally. At €160 million in revenue, close to 30% annual growth and approximately 6% net profitability, I suspect most readers here would be happy to own such a company in their portfolio.

// why am I sharing this?

Because Luca is an amazing business. And because I sometimes like to take a break from the AI buzz and see where entrepreneurship can take you in more traditional, less sexy industries.

The next exceptional business might be developing foundational AI infrastructure, or it might simply be selling pretzels 😅

Today, big part of my job is asking questions. But even before, many of my relationships were built around asking them. Funny enough, when meeting new people, they often walked away feeling we had a great conversation - even when they knew very little about me.

Being genuinely curious is memorable. People like being heard. They like being challenged a bit. They like feeling that someone is actually paying attention. And asking questions is probably one of the simplest ways to achieve that.

// applying it to fundraising

I’ve seen similar behavior with good fundraisers too.

During fundraising, some founders treat calls as a one-way presentation after brief intros. Most investors get comfortable and expect to spend 20-30 mins asking questions. Changing the dynamic and taking the investor out of their comfort zone can play in founders’ favor. A super easy way to do this is by just asking a question. Not the standard “what’s your ticket size?” type of question. More like:

• what personally gets you excited about a company?

• what’s an example of a founder you helped recently and how?

• what changed in your process in the last 12 months?

These questions do a few things: they make the conversation more engaging. They make the investor more present (people naturally do this when they are expected to say something). They tell the investor something about you - you’re curious, thoughtful, and evaluating them as much as they’re evaluating you. And investors hate being rejected by the founders at least as much as the other way around - so most of the times they’ll put in the effort for a good answer.

1-2 of these questions are usually enough - you don’t need a lot. Sometimes a “what do you think” at the end might work out well too - but that’s a clear sales closing tactic - so be careful not to come across as too salesy. Prolonging the call by overdoing it is not a good idea. Best founders can explain their company clearly and quickly, hence the calls don't extend too long.

And one brief additional note - leaving room for the investor to interrupt / ask their questions plays in your favor - good sales people always do it.

// generalizing

Regardless of the business, I always appreciate founders who have good interpersonal skills. Good listening and curiosity make people memorable. A fundraising call is still a conversation between two people - not just a transaction. It’s an opportunity to meet someone new and learn - not just to pitch.

I like AI. I use it a lot. Claude is great. The hype around it is funny and relatable at the same time. Productivity gains are real, although sometimes I wonder whether we’re becoming more productive, more lazy or just outsourcing thinking.

// intrigue 🤔

But lately I’ve noticed a trend I’m not so excited about: founders creating decks almost entirely using Claude. Don’t get me wrong - I think it’s great to use AI to create a deck faster. I’d probably do it too. But I’m opening decks where it’s obvious from page 1 that the document came almost directly from Claude. And that’s a pity, because somewhere in the process the deck loses a bit of the authenticity that should make founders stand out.

// the problem

Depending on the prompt, I think many AI-generated decks could be cut in half and then polished. Claude tends to create these very complete, corporate documents that mention every single thing from the prompt. The result is often a 20+ slide deck where everything is technically there, but somehow nothing stands out (not to mention AI mixing things up). So the issue is not creating faster with AI, but not touching the final results.

👉 What if you received a Europass CV for a top position in your firm?👈

// back to basics

I generally prefer simple decks. 6–10 slides are usually enough. Simple, clear, authentic (and ofc including the relevant data). Because the deck is not a compliance document.

// funny story

I even had one funny situation where a founder was presenting something from the deck and got confused about what was written there because it had been prepared very recently. It was an AI hallucination. Those obvious mistakes suddenly make the whole thing feel less credible.

I’ve also heard investors saying something along the lines of: if founders don’t spend time preparing the material they send, why should investors spend time reading it?

Maybe that’s too harsh. But I think there’s something underneath it that is true.

//conclusion

The pitch deck is the document that represents a founder’s life work. Founders should treat it accordingly.

Best founders do. Use AI. Generate the first version. But then remove things. Rewrite. Simplify. Make it sound like you. Or I don’t know, at least prompt differently than the rest of the founders around 🙂

Below is the data:

Within the CEE region, there are a total of apx. 640 data centers. Three countries are pulling ahead:

• Poland: ~93 data centers

• Turkey: ~82

• Romania: ~63

And these countries are actually important tech hubs in the region. I’d say the data did not surprise me.

I know there is a lot of interest to invest in data centers across the region. From what I gathered, half of the ongoing projects are in Romania (source).

Thoughts

Western - Eastern Europe

For context, Europe hosts about 3350 data centers (excl. Russia). The density is apx. 5.5 data centers per million of population. Eastern Europe hosts around 640 of them. That’s 5x less than Western Europe! Although the split of population among Western & Eastern Europe is approximately the same.

The digital infrastructure is lagging behind and that’s no news. Of course, the demand was not there - that’s the simple reason why the region is lagging behind.

But this gives support to arguments around data sovereignty and imbalance of how data is being used & stored.

I’ll look into how the data above correlates with the energy resources used in each country. Maybe that will bring interesting insights too.

US - Europe

US hosts apx. 3950 data centers. Compared to Europe’s 3350, the gap is not that large. I expected it to be much larger. The density is indeed double (11 data centers per million of population in US).

But keep in mind that many of the European ones (esp. in CEE) are smaller colocation centers and not hyperscalers. Hence, we should also look at the MW capacity to fairly compare the capacities of the different regions (not in this post though!).

All signs suggest that as the industry expands, CEE will increasingly attract data center investment.

The 💸 is there

There’s plenty of money in the market. VC funds have raised new capital, round sizes are up, valuations are climbing. But one type of capital seems to have quietly disappeared: experimental or education money (came up with this term while walking with a peer).

Most funds have successfully raised larger funds - everybody wants larger AUMs. That’s good news, but it also means they now need to deploy bigger tickets. And bigger tickets don’t work well at the pre-seed or acceleration stage. It’s not scalable to support hundreds of scrappy startups with small checks. Hence, few investors want to touch anything under €500-750K.

The bar is up

At the same time, the bar has gone way up, especially for first-time teams. Due to larger tickets, funds want conviction. They want proof. Which means they all end up chasing the same few teams that look “obvious.” That drives expectations even higher. Hence, we start seeing companies coming out of stealth raising €10M seed rounds (see Kontext case). We’re not yet such a mature ecosystem to have many of these.

But here’s the problem:

The small acceleration tickets played an important role in the last 8–10 years across CEE. €200K–€500K checks gave founders the chance to test ideas, fail, iterate, and learn. That money educated a generation of founders and built the pipeline that fed into recent times’ Seed and Series A in the ecosystem.

=> friction for newcomers

I think this scares away decent, not rockstar yet, founders that want to leave their job, build a team and start something new. And when they start, they might end up bootstrapping until they can show traction and then go straight to a Seed. This leads to less early deal flow and hence slimmer pipeline for seed and series A in the future.

There are some public (& private) initiatives again - which I applaud. But let’s be honest - we like capitalism, that’s why we’re investors. The market pushed us here, and we shouldn’t expect governments to solve this gap for us (cuz the market is actually us - it’s so few of us).

A year later, we have the 2024 dataset. The change? Not huge.

Key data take-aways

• 13.48% of EU enterprises (10+ employees as defined in the data used from Eurostat) used AI. That’s up from 8% in 2023 - a 5.5% jump

• 9% of Eastern European enterprises used AI compared to 5% in 2023

• Slovenia is still the leader (with 21%). It has the largest increase too - 9.5%

• Romania is lagging behind with 3% of enterprises using AI. Even though the growth is “staggering” - double compared to last year

• Turkey’s adoption churned (the only churn on the list) by 1%

Size, sectors and use-cases

The big-company bias is still large (more large enterprises use AI than smaller enterprises). 41.17% of large enterprises (250+ FTEs) use AI compared to only 11.21% of small enterprises (10–49 FTEs). So the AI revolution is still largely a “big balance sheet” game.

In terms of sectors - companies in information & communication sector use AI the most - 48.72% to be specific. That’s not a surprise as this includes the tech industry. Professional, scientific & technical services follow at 30.53%. Every other sector is below 16%.

Out of the companies using AI, 34% of them use it for marketing and sales, 27.5% use it for admin work and 22% use it for security.

Thoughts

It’s clear that:

1. Big enterprises in Western Europe or US adopt AI much faster (no news here)

2. There is still a lot of greenfield opportunity - 85% of companies in Europe do not use AI. Yet.

3. Penetration in large enterprises (3000+ FTEs) is strong. I believe 40% in Europe is quite good. And based on various research outlets, it seems like the number is pretty similar for large enterprises in the US. Does that show us that Europe is opening up?

// attitude & confidence

What’s super interesting to me when I talk to CEE founders going through YC is the shift in attitude they develop during the program. There are great upsides - like confidence. YC alums come out sounding like they can do it all. And honestly, confidence is something CEE founders could use more of.

// overconfidence

On the other hand, I also sometimes feel like they forget where they started. We all know things move fast in SF. Many investors, much easier to fundraise, great exposure, etc. But don’t disrespect investors for asking if, for instance, you have a budget. It’s okay if you don’t, but don’t become cocky and throw back a “have you ever invested in a YC company before?” Because many times, the answer is yes. I really appreciate founders who stay grounded. 3 months ago, these same questions wouldn’t have felt out of place.

// valuations

Regardless, one thing being discussed within CEE VC circles is the high valuations of the YC companies compared to local ones. I don’t think the gap is a big problem. It’s okay. But founders need to realize the pressure to deliver. Raising at $30M+ valuation out of YC is not out of the ordinary, so be prepared to show enough progress to defend a $60M+ valuation in 1.5 years. As long as you live up to the expectations, it’s all good.

And please don’t pretend like $30M is nothing, because it is not - even in the US. You probably know that there are some U.S. investors who go after the “5 at 25” rounds, but many who don’t.

// raising $ from home

It makes little sense for SF based founders to raise their entire rounds from home. This raises questions about a founder’s ability to attract international attention. However, a split between U.S.-based and home-based investors is great. There are various arguments for this, but if you raise nothing from home, you might struggle to tap into local talent networks (which everyone wants access to, esp. technical talent), and you could find yourself a bit isolated when that big U.S. name on your cap table doesn’t have the bandwidth to support you closely.

And yes, raising in SF can happen suuper fast. $3M+ in a matter of hours is not uncommon.

// final thought

But not all YC companies are successful - even if they fundraise easily. If we look at the numbers, there are higher chances of investing in a bad company from YC than a good one. It’s great to be part of the community, but stay humble. Many European investors can help. Success is much further away.

It’s clear that companies with global ambitions disregard the region’s stock exchanges altogether, heading to places like New York, London, or Frankfurt. But I decided to dig into the numbers to understand the regional dynamics. I think the stock market of a country is a variable that shows the investment landscape maturity of a country.

So here’s the data:

Warsaw: the undisputed leader

By far, the standout in the region is the Warsaw Stock Exchange (WSE). With the largest market capitalization in CEE, Warsaw also has a few IT companies listed. 23 out of about 400 to be more specific. What sets Warsaw apart is not just its size but its ability to attract companies from outside Poland itself. The WSE has established itself as a regional hub, offering international visibility and liquidity that other exchanges in the region simply cannot match.

It’s not surprising, then, that companies from countries like Ukraine, the Baltics, and even the Czech Republic have chosen Warsaw as their listing venue. This cross-border appeal cements its status as the go-to market for those aiming to break beyond their local ecosystems.

The surprise of Athens

What did surprise me, however, was the scale of the Athens Stock Exchange. It ranks second in market cap among CEE stock exchanges - an impressive feat for a market I hadn’t considered to be so significant. While Athens primarily serves Greek companies, its size demonstrates that even in markets with challenges, there’s substantial activity and a notable level of investor interest.

Bucharest: local but growing

Coming in third is Bucharest, which primarily caters to Romanian companies. Unlike Warsaw, the Bucharest Stock Exchange hasn’t yet positioned itself as a hub for the broader region. Most of its listings are domestic, and while it’s growing, its reach remains relatively constrained due to the traditional type of companies being listed (with local operations). Interestingly, the largest European IPO in 2023 (Hidroelectrica) was in Bucharest.

Observations

One thing that stands out across these markets is the type of companies they host. The vast majority of listings in Warsaw, Athens, and Bucharest come from traditional industries - banks, utilities, energy companies, and manufacturing firms dominate the rosters. While there are pockets of innovation, these markets are still waiting for their breakout tech champions to make waves.

Also, if we cross the pond to Istanbul (which typically falls outside the region commonly referred to as CEE, but still Southeast Europe), we find another sizeable stock market. With about €320B market cap, it’s right after Warsaw.

And yet, even when you add up the market capitalizations of all the stock exchanges in CEE (total of apx. €1.17T), the sum still falls short of the market cap of a single global tech giant like Meta. It’s a humbling comparison that underscores how far the region’s markets need to go to compete on a global scale.

So, if you had to list your company in CEE, where would you go?

1. Recap: local funding in Romania

The How to Web report shows a total funding volume of €130.7M in 2024 for Romanian startups. This figure covers everything from pre-seed to Series B+ rounds, plus various convertibles.

Worth noting is that about €35M are new rounds, while about €95.7M are follow-on rounds.

2. Updated data on diaspora

In my original article, I mentioned that Romanian diaspora founders had raised about $104M in pre-seed, seed, and Series A (notably from companies like Datology AI, Ezra, Tektonic AI, etc.). Now, let’s expand that to include the end of 2024 data and align with the broader coverage of Series B+ rounds, as done in the How to Web report.

Here are some important diaspora rounds:

• Databricks: $10B (Series J)

• Harmonic: $75M (Series A)

• Dexory: $80M (Series B)

• Starcloud: $24M (pre-seed + seed)

• Cascade AI: $3.5M

• others undisclosed (another about $14.5M)

If we simply lump it all together - including the Databricks outlier - the diaspora figure climbs to $10.3B+ (i.e., $10B from Databricks plus another $301M or so). Obviously, Databricks skews this number massively.

3. Making sense of the numbers

• With Databricks included, we’re looking at over $10B raised by diaspora founders, dwarfing the €130.7M in Romania.

• Excluding Databricks gives us about $301M total for diaspora - still 2.3x higher than the local figure (considering the eur - usd conversion too).

• Keep in mind that the €130.7M local figure also includes many follow-on rounds (like FintechOS’s €60M series B extension) - some of which started years ago - while much of the diaspora data I’ve listed above references relatively fresh raises.

Thoughts

Overall, I think the conclusion remains the same: diaspora founders manage to attract more capital largely because they have better access to funding networks and, in many cases, more advanced know-how derived from operating in mature ecosystems. In fact, I tend to believe that many experienced diaspora founders left Romania precisely because they struggled to find the right support - be it in terms of capital, mentorship, or market opportunities - back home. This doesn’t mean local founders can’t succeed; rather, it emphasizes the importance of building stronger support structures within Romania’s startup ecosystem, so that talented entrepreneurs feel they can thrive.

If you are a Romanian diaspora founder I haven’t mentioned in my data, let me know! I’m always refreshing these figures and would love to add more data points.

I kept my instructions very simple: I wanted an app where users could upload images and convert them into black-and-white cartoon versions - something kids could easily draw. That’s it. No complex prompts, no detailed wireframes. Just the basic idea.

Five minutes later, I had a fully functional app.

I’m talking about authentication, a paywall (as I chose to have a few free trials, of course), Stripe integration for payments, image conversion functionality and even a name and logo (without me even asking for it!). The app was ready to deploy, and the AI walked me through the code. If I need any modifications, all I have to do is ask, and it adjusts accordingly. It even explains why some features are worth implementing in a way or another.

Price

The total cost of all mentioned above? $0. But eventually I, of course, hit a paywall. To continue to work on this I have to pay $15/month (on a yearly plan), and if cloud usage increases, the cost scales up. But let’s be real - $15 is peanuts! You can’t talk to someone with this money about this app, let alone build anything for you. And here I was, with a working prototype in minutes. Btw, the conversion actually works, I just have to play a bit with it to improve the output.

Final thoughts

This is the kind of thing that makes the kid in me want to play around endlessly. While tools like Bubble have made no-code development more accessible, AI takes it to another level. It’s not just about building - it’s about having an intelligent assistant guiding you every step of the way.

That said, while AI can handle simple app-building tasks like mine, it’s important to recognize that senior engineers, architects, and experienced developers are still very much needed. AI can streamline the process for smaller projects and individual creators, but large-scale applications require deeper involvement, planning, and expertise that I think AI alone cannot yet provide. Rather than replacing developers, AI is more likely to shift their focus to more complex, high-value problems while automating repetitive tasks.

For anyone with an idea, you should go explore. AI isn’t just changing the way we build - it’s changing who can build.

Fundraising can be one of the most pivotal (and stressful) moments in a founder's journey. Done right, it opens doors to growth and opportunity. Done wrong, it can signal red flags to investors and close doors before you even get started. Over the years, I’ve noticed two recurring mistakes that early-stage founders make during fundraising. These missteps are entirely avoidable, yet they’re surprisingly common especially for founders new to fundraising (& in CEE).

Let’s break them down.

1. Asking investors how much you should raise

One red flag is founders asking investors, “How much do you think I should raise?” when pitching. This signals two things: a lack of preparation and a lack of confidence.

As the founder, you know your business best. You understand your product roadmap, your hiring needs, and how much capital it will take to hit your next big milestones. If you’re asking investors to make that call for you, it suggests you haven’t done the homework.

A similar variation of this mistake is when founders say something like, “We’re thinking of raising at least this much, but it depends on how you see things.” While it might sound collaborative, this approach still signals uncertainty. Investors don’t want to feel like they’re guiding the ship - they want to see that you’ve mapped out the course and are confident in your plan.

So, do your homework first. Work through the numbers, draft a clear plan, and, most importantly, get feedback—whether that’s from mentors, advisors, or even trusted investors before you start pitching. If you’re not prepared or still feel unsure, it’s far better to postpone the meeting. Showing up unprepared can leave a lasting impression of uncertainty, which can be difficult to undo. A delay is temporary, but a weak first impression can be permanent.

What investors do appreciate is a clear, well-thought-out plan. Show them the logic behind your ask: how much you’re raising, how long that runway will last, and what you’ll accomplish during that period. If they see gaps in your plan or ways to optimize it, they’ll share their feedback. But don’t expect them to create the plan for you—that’s your job as a founder.

2. Name-dropping other investors without substance

In a world where momentum matters, founders often think name-dropping other investors can create FOMO. You might say, “I’m already in talks with X, Y and Z investors” The goal, of course, is to add credibility and make the opportunity seem urgent. But here’s the catch: it’s a very small world.

Investors talk. They know each other. And it’s easier than you think for them to validate whether you’re truly in advanced discussions—or just trying to bluff your way into their checkbooks. Once they find out you just sent a deck to that investor, your credibility takes a nosedive. Trust is fragile in fundraising, and the last thing you want is to start your relationship with an investor on shaky ground.

Tip: Instead of overhyping non-existent momentum, focus on being transparent about where you are in the process. If you don’t have term sheets yet, emphasize your progress, traction, and vision instead of leaning on the illusion of external validation. Authenticity goes a long way.

Final thoughts: build trust, not tricks

Fundraising isn’t just about the money—it’s about building relationships. Investors want to back founders who are thoughtful, prepared, and trustworthy. Falling into the traps of lack of planning or name-dropping might feel like shortcuts, but they almost always backfire.

Instead, focus on doing the work. Know your numbers, craft your story, and approach investors with confidence and integrity. Momentum is powerful, but it’s best when built on substance, not smoke and mirrors.

If you’re a founder gearing up for fundraising, take a step back and ask yourself: Am I ready to inspire trust? Because that’s the foundation of every successful funding round.

Revisiting the Numbers

For consistency, I used the same assumptions from my last analysis:

• Conversation Speed: 60 words per minute.

• Word-to-Token Ratio: 1 word equals 1.3 tokens.

• Continuous Interaction: 24 hours a day, every day.

This means a daily token count of: 60 words/min × 60 min/hr × 24 hr/day × 1.3 tokens/word = 112,320 tokens/day.

Monthly token count: 112,320 tokens/day × 30 days = 3,369,600 tokens/month.

Let’s crunch the numbers with DeepSeek’s pricing tiers and OpenAI’s most advanced model.

1. DeepSeek’s Pricing

DeepSeek offers the Reasoner model:

• DeepSeek-Reasoner:

  • Input tokens: $0.55 per 1M tokens.

  • Output tokens: $2.19 per 1M tokens.

For simplicity, I’ll assume a 3:1 ratio of input to output tokens, a common pattern in AI usage.

Cost Breakdown for DeepSeek-Reasoner:

• Input Tokens: 3,369,600 tokens × (3/4) = 2,527,200 tokens.

• Output Tokens: 3,369,600 tokens × (1/4) = 842,400 tokens.

• Input Cost: (2,527,200 / 1,000,000) × 0.55 = $1.39

• Output Cost: (842,400 / 1,000,000) × 2.19 = $1.85

Total Cost for DeepSeek-Reasoner: $1.39 + $1.85 = $3.24/month.

2. OpenAI’s Most Advanced Model

OpenAI’s most advanced model, o1, offers the following pricing:

• Input tokens: $15.00 per 1M tokens.

• Cached input tokens: $7.50 per 1M tokens.

• Output tokens: $60.00 per 1M tokens.

Cost Breakdown for OpenAI o1:

• Input Tokens: 3,369,600 tokens × (3/4) = 2,527,200 tokens.

• Output Tokens: 3,369,600 tokens × (1/4) = 842,400 tokens.

• Input Cost: (2,527,200 / 1,000,000) × 15.00 = $37.91

• Output Cost: (842,400 / 1,000,000) × 60.00 = $50.54

Total Cost for OpenAI o1: $37.91 + $50.54 = $88.45/month.

Observations

• 27x price difference;

• OpenAI updated its pricing since last month. In the last post, the estimated cost was about $404 / month. So 125x price difference here.

While DeepSeek is the obvious choice from a cost perspective (probably performance too, but I guess still to be validated), I wonder how strong the brand value of OpenAI or other big providers actually is. I have doubts companies will start deploying from a new Chinese vendor right away due to so many reasons. We still have to see if DeepSeek will be able to cross this skepticism bridge.

Closing Thoughts

In my previous post I stated we’re in a race to the bottom. Well, it’s still going on hard! I guess this was pretty foreseeable, but soon enough we’ll wonder how sustainable this is. I believe costs will continue to decrease.

In this case, will startups still need to stress about ways to adapt to this new COGS line in their P&L (e.g. founders investing in their own mini data centers)?

And of course - if sales don’t work, investors won’t invest unless there are good reasons for it (e.g. business model, R&D heavy product, etc.). Showing early signs of success signals potential of more success in the future. Early signs of being unreasonable do the exact opposite.

How sales teams are naturally built

“We just need to hire sales people now” can be such a bad thing to hear from very early stage founders. Going out and hiring some sales reps won’t solve the issue. Founders are the ones who truly understand the product and customer pain points, which are essential to driving the sales strategy. Especially early on when there is no script and you have to be flexible & creative. When the process becomes repeatable and you start seeing consistent results, that’s when bringing in additional salespeople can help scale things. The founders can then set the example and establish the sales motions that others can follow. Once the team starts seeing consistent success with 2-3 sales reps, it's time to bring in a senior sales executive to lead the team effectively, but only if the foundation is strong.

Bringing in a senior leader too early could hinder flexibility and momentum, as they might focus on imposing structure rather than adapting to the startup's needs. Additionally, senior execs cost a lot and do not necessarily sell. They build teams that sell.

Industry know-how is so important for sales leaders

Working with a sales executive who has a stellar background but doesn’t understand the specific nuances of your industry can kill companies. I’ve seen this in the past. They might miss important market insights or fail to grasp the unique needs of your customers, leading to missed opportunities. This lack of understanding can cause costly delays as they try to get up to speed. Getting up to speed fast can happen, but they still need to be flexible and hungry to sell.

Sales cures all (incl. fundraising)

The foundation of any successful business is built on the ability to sell. Investors want to see signs of success, and consistent sales can provide that. If you add a great product and team, what else can you ask for? So before looking for outside funding, make sure your sales process is solid and repeatable, and don’t rush into hiring an executive team before you’re ready.

Enjoy the holidays! Hope Santa brings the gifts you wished for!

Breaking down the numbers

I started with the cheapest model out there - Together AI’s lowest pricing tier is $0.06 per 1 million tokens. Here’s how the numbers add up:

1. Together AI’s Pricing

• Assumptions:

  • Conversation speed: 60 words per minute.

  • Word-to-token ratio: 1 word equals 1.3 tokens.

  • Continuous interaction: 24 hours a day, every day.

• Daily token count:
  60 words/min × 60 min/hr × 24 hr / day × 1.3 tokens / word = 112 320 tokens/day

• Monthly token count:
  112 320  tokens/day × 30 days = 3 369 600 tokens / month

• Monthly cost:

  (3 369 600/1 000 000) × 0.06 = 0.202 $/month

For $0.20 per month, you could talk to Together AI’s model non-stop, 24/7. Pretty cheap, isn’t it?

2. OpenAI’s most expensive model

Now, let’s contrast that with OpenAI’s most expensive pricing tier: $120 per 1 million tokens.

• Monthly Token Count (Same as Above): 3,369,600 tokens.

• Monthly Cost: 3 369 600 / 1 000 000 × 120 = 404.35 $/month

At OpenAI’s rates, a similar 24/7 usage would cost $404.35 per month. That’s 2,021 times more expensive than Together AI’s lowest pricing tier.

Most of the other vendors fall somewhere in between these final costs.

Costs improvement from data curation

Just before posting this I saw this post from Datology (startup developing data curation technology for optimizing the use of AI). RO founder btw!

A new frontier in AI efficiency lies in data curation. By curating high-quality training data, they’ve shown remarkable gains in training efficiency, performance, and inference costs.

Highlights from Datology’s results:

1. Train faster: By curating the training data, they achieved the same baseline performance 7.7x faster. This drastically reduces the cost of obtaining results and speeds up iteration cycles;

2. Train better: DatologyAI’s curated dataset improved performance by 8.5 absolute percentage points over the baseline (60.5% vs. 52.0%);

3. Train smaller: With better data, smaller models could be trained to outperform larger ones, cutting inference costs by 2.1x while still improving performance over the baseline by 5.7%.

These optimizations reduce the operational costs of AI significantly. For businesses training or customizing their own models, these savings in time, cost, and performance are game-changing.

What this means for the AI landscape

This comparison highlights the race to the bottom in AI pricing. While Together AI charges a fraction of a cent for tokens, OpenAI’s pricing—while justified by model quality and features—caters to a different market tier.

Key implications:

1. Cost vs. Quality: OpenAI’s higher pricing reflects the advanced capabilities of its models, which may be indispensable for enterprises with high-stakes use cases. Together AI, meanwhile, offers a highly competitive option for more basic or volume-based use cases.

2. Accessibility: Together AI’s low pricing opens doors for startups, small businesses, and hobbyists to explore AI without worrying about costs.

Reflecting on unit economics

This exercise ties directly back to my earlier article on startups’ unit economics. If AI becomes this cheap to operate at scale, it’s a game-changer for businesses relying on AI-driven services. Customer service, product recommendations, and even personal assistants could become standard offerings, as costs would no longer be a barrier.

Yet, this also poses questions:

• Will providers like Together AI rely on massive adoption to sustain such low pricing?

• Can premium providers like OpenAI justify their pricing over time, or will they also need to adapt to the downward pricing pressure?

Closing thoughts

The math here is pretty interesting in my opinion. On Together AI’s pricing, 24/7 interaction costs just $0.20 per month, compared to $404.35 per month with OpenAI’s most expensive model. This underscores how competitive the AI space has become.

The race to the bottom is real, and I am pretty sure prices will continue to decrease.

UI/UX in products

When it comes to user experience, simplicity delivered smoothly can elevate a product from being just “good” to “10x better.” This is especially true for B2C and product-led growth (PLG) businesses, where seamless design isn’t just a nice-to-have; it’s a competitive edge. A clunky, complex product, no matter how feature-rich, can alienate users. Beautiful branding and a well-thought-out interface make products not only more enjoyable, but also easier to adopt.

UI/UX for investors

The importance of good design doesn’t stop with customers though - it extends to investors, too. Think about the first impression a startup gives during a pitch. Two companies with identical metrics but different levels of design polish will likely have different outcomes. A well-designed pitch deck, a user-friendly website, and even a thoughtfully organized data room signal diligence and attention to detail. These elements show investors that the team can build something customers want & can market accordingly.

Additionally, complete & beautiful commercial data rooms (or documents included in them) help in building up momentum. They limit friction (going back & forth asking for docs, answering questions) and puts you in control of the process. All investors have the same information and depending how you time your process, you can infer more about their interest, process, speed.

Investors are human, and like everyone else, they’re influenced by aesthetics, often unconsciously. We see hundreds of decks and websites, and when a company’s design aligns with what we associate with success, the impression sticks. Not all companies take advantage of this, so if you do, you’ll set yourself apart. And you’re more likely to get the following reaction:

A call for beauty

As a region, we’re strong technically, but we need to be “beautiful”, too. This isn’t just about vanity; it’s about creating trust, signaling competence, and resonating with both users and investors. Look at any startup you admire - branding is one of the first things they prioritize as they scale.

Humans are naturally drawn to beauty. We should leverage that bias, not just to attract attention, but to showcase the strength and potential behind our work. Great teams build great products together with beautiful support - decks, presentations, websites, data rooms. So please, take some time to make it beautiful!

Maybe your focus is not there.

Why focus

To do well, you need one very good solution that is 10x better than competitors’ for one specific group of customers. You don’t need 5 products that are each 2x better. Have too many nice to haves and you’ll dilute the core. That applies to early stage. Sure, once you’re big enough you’ll be able to diversify in some many areas. Think Amazon - starting with selling books online but now providing cloud services.

And when I say focus, I also mean industries you’re after, countries you’re targeting, go-to market motion. It’s not only about the product. I would actually argue it’s more about sales & marketing. Focusing on a specific customer segment leads to faster feedback loops, stronger customer relationships and tailored approaches that resonate with them.

Competition is so tough these days. And as an early stage founding team you have limited resources. That’s why you have to focus and be the best in class at your one single thing for which you can show efficient strong traction.

Limited resources

If you go too wide, you’re going to lose valuable resources - time & money. Just as a personal example - at SeedBlink, one of the best things we did at some point was to refocus. Even sales people did not know what they were selling anymore because we were doing too many things at once too early. And I’ve seen that with other companies, too.

When you’re at pre-seed, seed & series A you you just don’t have the luxury to make many mistakes. Clock is ticking and if your sales rep who you spent 2 months hiring & 2 months training does not catch any of the 5 rabbits he’s after, you’re already in a tough position. Multiply that with a few other hires you made and it’s worse.

Culture & fundraising

That does not only impact your burn and metrics, but has a large impact on the company culture too. Too wide can make employees collaborate less and when things don’t work out you’ll see them pointing fingers at each other. Productivity goes down.

Focus and simplicity will also help you fundraise. Not only that investors will quickly understand your business (which will allow you to get into deeper conversations faster), but you’ll most probably have better numbers and show you’re a good planner, too.

Case studies of successful, focused companies

• Amazon started by selling books online but has since expanded into a global e-commerce giant and cloud services provider;

• Slack launched as a messaging tool for small tech teams but is now widely used across industries;

• Facebook began exclusively for college students, but now connects billions of people worldwide;

• Netflix initially focused on DVD rentals by mail but transformed into a streaming platform and content creator;

• Airbnb started by renting out air mattresses in single apartments, but now facilitates stays worldwide, from shared rooms to luxury villas;

• etc.

Final note

Focus is key, but customers always lead the way. Listen and iterate based on their feedback. And ask yourself: “Do we have a clear target customer? Are we trying to do too much too soon?”

That's because communication is an important skill in building a business. To me, the way founders communicate often reflects how they will grow an organization- hiring & motivating people, selling globally, building a culture, opening up offices, etc. This is a big part of what VCs bet on: founders who can deliver not just on metrics but also on building an organization.

It’s clear that first-time founders do not have all this experience, and they need time to grow into the role. Investors, however, seek founders with the skill sets that maximize the chances of evolving successfully. It is debatable which skills matter most, and each investor has their own perspective.

But one thing I want to highlight is the importance of being able to communicate effectively in an internationally recognized language (aka English). It should be taken for granted, but it sometimes isn’t. How can you do all the above if speaking English is already out of your comfort zone?

I've seen this scenario many times recently and it is honestly a waste of effort for founders. Putting together a team, a product, a data room, etc. takes real effort! Yet, if you can't communicate effectively about them, it's hard for anyone to take you seriously.

Lacking communication skills (incl. English proficiency), can lead to a lot of frustration and stress. Founders may feel neglected, and they might not understand why they're being passed over by investors. This can build up into resentment, making founders cynical about VCs and the ecosystem as a whole.

I believe that over time this won't be a topic of discussion anymore. Yet, in the meantime, I strongly recommend founders to work on perfecting their English before jumping into building a tech startup. Taking rhetoric or public speaking courses could also make a big difference in the long run.

The effort founders put into becoming better communicators will pay off, not just with investors but with their teams, partners, and customers too. I’m always excited (& I am sure others are too) to meet great communicators.

I grew up in Romania and I always had the feeling we’re super mercantile and we like to find our own ways to do things. We’re entrepreneurial. Just last week I literally met a grandma of a friend who recently went abroad for a trip and came back with fragrances and other random chippies to sell to her neighbors or colleagues. For some, it’s in our blood to do it. I myself was selling food (and others) in high school.

Ever since the past communist regimes, there has been a generation of new born entrepreneurs selling ANYTHING like crazy - clothes, gold, etc. That was the case for Romania, Poland, Czech Republic and others in CEE. Many of them built large businesses locally. We now have so many examples of traditional businesses around us that do well - hospitality, distribution, logistics, etc. In many cases, even politicians do business! ;)

And I thought about this because Istanbul is, of course, the home of this spirit - that’s where founders in the 90’s were going for their supplies, right?

Why don’t we see the same attitude when it comes to selling software products?

I was thinking about how startups in RO & CEE are aligning to this vision. Well, most of the time they don’t.

So often I see teams prioritize products and not sales. Most of the startups are failing because the product could not be sold. What if they failed because the product was not delivering to its promises? Well, I’d prefer that.

Many times I feel like some startups forget about the main goal of a business - delivering value & making money. They many times put product, fundraising or others ahead of this. Launch’s startups annual report in 2023 even shows that founders see one of their biggest challenge being “sales”. I was amazed - how can that be? If sales is the biggest challenge, then the business cancels itself. It’s like a supercar that does not work - is that valuable anymore?

Go ahead and share this article to your startup friends!

Share

We need more sales hustling

We need the same sales / entrepreneurial spirit from traditional businesses - where you can’t rely on VC’s money - in todays’ startups. I agree, the context is completely different. However, technology should empower entrepreneurs and improve their productivity, right?

Having the clients is such a huge asset! Great product builders could then deliver a differentiated product for the clients. For a good startup, for sure you need both - but having the first one makes everything else flow so easily!

Embracing the mercantile mindset

We need this mercantile mindset in most startups. GTM is about testing the waters, finding those early adopters, and pivoting quickly to meet their needs. It’s about being on the ground, getting feedback, and evolving faster than anyone else. A good enough product with a clear market fit can win the race against a perfect product that few people want.

I believe this shift in mindset could help grow the ecosystem faster, leading to incremental successes that pave the way for future unicorns. This approach may apply mainly to application-layer businesses, while R&D-heavy products could be different. Still, any business at any stage has traction / validation points.

And you know what’s the best part? It’s already in us.

I’m thinking about ways to instill this in the ecosystem - be it with content, workshops or others. So if you have input, please do share!