Inception Fund

Still in pre-training. Backing New Nordic made technical brilliance at the earliest inception stages.

www.inceptionfund.vc

In pre-training...

Zenith Ventures

Founder and operator turned VC. Backing obsessed Nordic B2B software and vertical AI founders from the earliest stages. 

No-nonsense, hands-on and fighting for founders.

General Partner

Stealth AI Startup

Pre-seed Investor

Natively

Build and launch native mobile apps with AI.

www.natively.dev

Lovable

Building the last piece of software.

Careers: https://lovable.dev/careers

www.lovable.dev

Angel Investor

Vantel (YC W25)

AI Software for commercial insurance brokers (YC W25).

www.vantel.com

Pre-seed Investor 

Tandem Health

Helping clinicians focus on providing care instead of on administrative tasks through AI and automation.

Careers: https://jobs.ashbyhq.com/tandem-health 

www.tandemhealth.ai

Tzafon

An applied AI research & development firm looking to push the boundaries of what machines are capable of through the intersection of AI & software engineering.

www.tzafon.ai

Lightbringer

Navigate and protect your IP and patent journey with AI.

Careers: https://www.lightbringer.com/apply

www.lightbringer.com

Seed Investor

Pruna AI

The AI optimization engine for ML teams seeking to simplify scalable inference.

Careers: https://careers.pruna.ai/

www.pruna.ai

Data & AI/ML Entrepreneur | Early Stage Investor

Over a decade of experience within Big Tech, B2B software, Data and AI Infrastructure as a founder, operator and investor.

Passionate about bringing new and paradigm-shifting technologies to market. This passion has led me to a over decade-long entrepreneurial career focused on emerging software industries and global markets - from ML & Data Infrastructure startups and venture capital funds to Fortune 500 technology leaders such as Google. 

Raised +$50M in early-stage funding from top VCs as a founder and operator. 

A big believer in people, data, design and telling a story. Passionate about building new products with the right GTM, messaging and growing teams that do so.

Deep understanding of the venture capital, private equity, SaaS, cloud, data and AI ecosystems with solid global relationships at all levels.

Frequent keynote speaker on topics around data, AI, startups and venture capital. Considered as one of the top AI & data thought leaders in the Nordics.

I don't think people truly understand how early we're in enterprise adoption of AI agents.

McKinsey just released their 2025 AI report. 

Some nuggets:

1️⃣ 90% of companies “use AI,” but 67% are still stuck in pilot mode. Corporate AI theater is alive and well.

2️⃣ 62% of orgs are experimenting with AI agents, 23% are scaling AI agents. Most are in tech and healthcare.

3️⃣ The impact gap is massive. 64% say AI helps innovation, but only 39% see real EBIT gains. 

4️⃣ The high performers (top 6%) think bigger. They rebuild workflows, set growth goals, and invest real budgets not just PoCs.

5️⃣ Leaders who own AI personally are 3x more likely to scale it. Makes sense.

6️⃣ The winners use AI to transform how work gets done, not just speed it up.

7️⃣ The average company measures efficiency. The best ones measure how fast their agents can act.

8️⃣ Risk management is catching up with 51% have already seen AI backfire, mostly from inaccuracy.

9️⃣ The workforce impact is foggy. 32% expect cuts, 13% expect growth, everyone else is guessing.

Agentic AI looks magical in demos, but scaling it across teams and systems (especially in enterprise settings) is where it still breaks. You can do basic single-step automation reliably right now (tool calling).

As Andrej Karpathy recently noted on LLM agents:

"My critique of the industry is more in overshooting the tooling w.r.t. present capability."

Bottom line: we're still super early, this transformation will go on for another decade.

I'm bullish, but this will take a bit longer than we think, having fully autonomous agents. Just like autonomous driving took a bit longer than we thought, but here we are.

E.g. one of the biggest challenges in building AI agents today is agents' long-term memory.

#artificialintelligence

Oliver Molander

What tech stacks are top early-stage AI startups actually using? 🤔 

Below are some of the data points I've gathered after speaking with hundreds of AI startups.

1️⃣ Frontend:

Next.js dominates across almost all teams; Tailwind CSS is the standard choice for styling; ShadCN UI components are somewhat common; some use Chakra UI; TypeScript is universal (no exceptions).

2️⃣ Backend:

Most teams use Python (typically with FastAPI); Node.js + Express remains popular for lightweight services; Go is used selectively for performance-critical microservices; Postgres is the main database; Widespread use of SQL + vector DBs, with pgvector dominating (I'm a bit worried for standalone vector DB startups).

3️⃣ Deployment:

AWS and GCP remain most popular, BUT Azure is seeing a notable increase (especially compared to post-ChatGPT); Many combine Vercel for the frontend with AWS/GCP for backend infrastructure; GitHub Actions is the most common CI/CD choice; Docker is standard (Kubernetes is used more selectively); Serverless adoption is rising (with growing use of e.g. Modal for on-demand compute).

4️⃣ AI Model Usage:

In proprietary APIs OpenAI/GPT, Anthropic/Claude, and Google/Gemini dominate; In open-source not seeing e.g. Meta or Mistral almost anywhere, Qwen is rising fast.

Almost no startups train models from scratch, and very few fine-tune. The real constraint is data, not model access.

Success increasingly depends on:

➡️ Backend optimization
➡️ Data structuring and formatting
➡️ Integration of AI into business processes

Roughly 90% of effective AI implementation is data engineering and operational design, not model architecture.

Fine-tuning should be a last resort. Before committing to it, most successful teams first:

➡️ Improve prompts
➡️ Add post-processing scripts
➡️ Fix their data pipelines

E.g. one startup reduced 90% of model errors with a simple script. Fine-tuning adds long-term maintenance overhead and should be reserved for cases where every other lever has been optimized.

Some leading AI companies are moving toward more efficient systems languages:

➡️ Lovable migrated 42,000 lines of Python to Go, improving concurrency, deployment speed, and maintainability.

➡️ OpenAI transitioned its Codex CLI (now ChatGPT CLI) from Node.js to Rust, prioritizing performance and reliability.

OpenAI’s codebase is known to be a large monorepo, primarily written in Python, with increasing usage of Rust and Go. It combines scalable infrastructure built by seasoned engineers with research-grade code from PhDs. Most APIs are built around FastAPI with Pydantic for validation, but style consistency varies widely.

Key takeaway:

The differentiator in AI startups is no longer model access - it’s data quality, backend optimization, and system architecture.

The teams that invest early in clean infrastructure and efficient data pipelines will outperform those focused solely on fine-tuning or proprietary models.

Thoughts?

#artificialintelligence

Almost a decade of experience within Data and AI/ML Infrastructure as a founder, operator and investor.

Passionate about bringing new and paradigm-shifting technologies to market. This passion has led me to a over decade-long entrepreneurial career focused on emerging software industries and global markets - from ML & Data Infrastructure startups and venture capital to Fortune 500 technology leaders such as Google. 

Raised +$50M in early-stage funding from tier 1 VCs as a founder and operator.

A big believer in people, data, design and telling a story. Passionate about building new products with the right GTM, messaging and growing teams that do so.

Deep understanding of the startup, venture capital, SaaS, cloud, data and AI/ML ecosystems with solid global relationships at all levels.

Still in pre-training...

Charles C.

Reverse founder factories 👇

Startups attracting founders.

We're seeing more and more of this, also in Stockholm.

And it's not zero-sum - we're also seeing an increasing amount of next-gen founder factories. The same companies attracting founders are also nurturing new founders.

And most importantly, there's no one right way to build an entrepreneurial career.

You can start a company. 

Then decide to join a hyper-growth startup. 

Then start a company again.

Never place yourself in a pre-defined box.

Play the game at the field.

#artificialintelligence

I think this captures the giving-back culture in SF quite well:

Recently a Nordic founder under 25 got into YC and decided to cold-email Sam Altman to ask him for lunch.

Sam replied and they had lunch.

This is probably the busiest CEO in the world right now, pulled in every direction.

Yet he made the time.

There’s something for European tech ecosystems to learn here in terms of giving back.

There’s also a pattern I keep seeing in young founders who eventually become successful:

They’re not afraid to cold-email people they admire and can learn from - and they keep it short, well-crafted, and concise, which leads to surprisingly good response rates (whether yes or no).

As Harry Stebbings often points out, the best founders are exceptional at cold emailing. A great, well-researched, articulate, concise cold email shows just as much hustle - if not more - than a warm intro.

When I was a little over 20, I cold emailed one of Finland’s most successful entrepreneurs (who was inte middle of a M&A transaction) to interview him for my master’s thesis.

He replied within three hours, and I interviewed him the next week.

I've applied that mindset throughout my entrepreneurial career for a decade plus.

Just go for it. 

A quick “no” is always better than a slow “maybe” (or no reply at all).

Keep it short, keep it concise and create momentum.

And most importantly, never take anything personally.

#startups

The Slush team still deserves a Cannes Lion award for this 🎯

See you in Helsinki next week badasses, my hometown 🥶

#slush

Super proud of the work that the NestAI team and Peter are doing for Finland - now together with Nokia and Tesi 🇫🇮🙌

Change is in the air.

#artificialintelligence

Super humbled to share the stage yesterday at the AI Summit with so many people from the startup, corporate, research and political world that I deeply respect 🙏  

And I want to highlight the work that Peter, ELLIS Institute Finland and PostScriptum do for Europe, the Nordics and Finland in terms of lifting the ambition level and attracting talent.

I moderated a panel together with Ilkka, Kristo, Alexandra and Nina - where we discussed the challenges and opportunities in connecting science and society, especially during the current AI boom and what many refer to as the space race of our generation. 

It was very interesting to hear opinions from different vantage points, given the different backgrounds of the panelists - from decacorn founders and European startup champions to technical enterprise leaders and science-based deep tech investors.

Here are a few nuggets from the panel:

➡️ Everything moves faster with AI, including the conversion of scientific discoveries into societal and business applications - we must adapt accordingly to compete on a global stage.

➡️ It’s also important that we change the negative narrative in Europe. Things are changing fast, and we have so many great AI companies emerging from ElevenLabs to Black Forest Labs, Syntheisa, Lovable, Legora, and many many more.

➡️ We should also dream bigger and take bigger bets in Europe. Here also connecting science and society plays a crucial part. European researchers have played key roles in groundbreaking innovations such as CRISPR, mRNA vaccines, and even the invention of the World Wide Web itself. 

➡️ In terms of timelines, startups think in days or weeks - while science and research usually operate in years - we need to find ways to mitigate this gap efficiently and decrease the cultural distance between startups and the science/research community.

➡️ When backing science and research teams, we need to blend in “startup hustle” faster into the founding teams at inception. Just look at OpenAI with Ilya Sutskever and Andrej Karpathy, who as research rockstars were founding team members together with Sam Altman, who came directly from the startup world and Y Combinator.

➡️ We need to be better at celebrating scientists who become entrepreneurs. In the US, they’re perceived as rockstars. Take Jussi Westergren as a concrete example - very few know he played a pivotal role during the early founding days of DeepMind (and he is a Finnish research entrepreneur that too few know about).

These were just a couple of examples.

In short: the opportunity is huge, but there’s plenty of work to be done to convert science into society and the business/startup world - resulting in economic dynamism. 

But we should be optimistic, especially in Finland and the New Nordics, given the research and engineering talent that we have here.

Let’s dream bigger.

#artificialintelligence

Two things I hate that inception stage investors do:

➡️ Asking “Why can’t a hyperscaler or AI lab (Google/Microsoft/Amazon/OpenAI/Anthropic) just do this?”

This question is especially dangerous in today’s AI super cycle. Even when the logic says a hyperscaler should do something, the reality is they usually can’t - because they lack the time, resources, or focus to solve every problem well. We’ve already seen this play out in AI code generation, where everyone assumed Microsoft Copilot had the market locked down a few years ago.

➡️ Questioning market size

Market sizing at pre-seed almost always leads to the wrong decision. The best founders either wield new markets into existence or dramatically expand existing ones. A classic example is Uber: most early investors passed because they thought the premium black-car market was too small for a venture-scale outcome.

There are countless examples like this. 

Even Google’s earliest investors assumed that if the company executed flawlessly as the twelfth search engine on the market, it might reach a ~$1B valuation at best.

At the very earliest stages, only two things matter:

1️⃣ The founding team - their technical strength and their ability to attract A-class talent.

2️⃣ Whether their initial idea is directionally correct, knowing it will drastically evolve.

Ideas and markets change.

People don’t.

In early-stage venture, the error of omission is far worse than the error of commission.

#venturecapital

I'm looking forward to moderating a panel with Alexandra, Nina, Kristo and Ilkka next week at the AI Summit in Turku 🙌

The topic is “Connecting science and society.”

Did you know that Finland is behind many scientific and research breakthroughs that have been successfully translated into society and the business world?

Such as Xylitol (developed at Turku University in the 1970s).

Finnish engineer Matti Makkonen is often called the “Father of SMS” because he pitched the idea of sending short text messages via mobile networks in 1984. 

We have a strong engineering culture in Finland. E.g. in the 1940s, a Finn built the first functioning black-box flight recorder. In the 70s, we invented the heart-rate monitor, and in the 90s we made the world’s first GSM call.

Who invented backpropagation? Its modern form (also called the reverse mode of automatic differentiation) was first published in 1970 by Finnish master’s student Seppo Linnainmaa.

These are just a few examples.

Linus Torvalds was a computer science student at the University of Helsinki when he began developing Linux (1991). The kernel itself was a student-led, open-source software project - not a formal research output. However, it emerged within a scientific environment: the culture of UNIX systems, computer-science education, and the open academic ethos.

Without Linux, the web as we know it would not scale.

Speaking of the web - did you know that the first graphical web browser, Erwise, launched in 1992 (before Mosaic and Netscape), was created by four Finnish students at the Helsinki University of Technology (now Aalto University): Kim Nyberg, Kari Sydänmaanlakka, Teemu Rantanen and Kati Suominen? 

Their supervisor was Ari Lemmke, the same person who gave Linus Torvalds’ “Linux” its name on the FTP server.

With the rapid rise of AI, it's more important than ever that we strengthen the bridges between science/research and society/the business world.

Super proud of the work that ELLIS Institute Finland and Peter Sarlin are doing.

Now more than ever, we should dare to dream bigger.

#artificialintelligence

I've changed my mind: 

ARR is still alive 🤫 

Just spoke in Stockholm with a founder who has 10xed revenues during the past 12 months.

Nothing special, you say? There are plenty of explosive AI companies from Stockholm 10xing their ARR right now.

Well, here comes the best part. 

This is REAL annual recurring revenue:

➡️ Signing seven-figure (euro) multi-year deals even with upfront payment
➡️ Closing some of the most demanding enterprise customers out there (e.g. banks)
➡️ Incredibly sticky product with strong expansion metrics and strong gross margins
➡️ Cash-flow positive now, so default alive and can decide on their own future
➡️ Has won famous international logos against US competitors who have raised ten times more in funding
➡️ Still founder-led sales - now with a scalable enterprise sales playbook (this is a technical founder who became extremely strong at enterprise sales over time)
➡️ Will clearly breach 8-figures (euro) in ARR next year - even without any additional funding (the company has raised a pre-seed and seed round)

This is the real deal.

I was originally taught that ARR is the value of the currently paying, deployed, and contracted customers, annualized. In other words, ARR is a guaranteed annuity. The guarantee implies predictability, which investors prize and value at high multiples.

Over time, the definition of ARR has slackened - especially with the GenAI boom. 

Feels like everyone are somewhat confused at the moment.

First, SaaS companies modified it to include customers who weren’t contractually obligated to pay, but did pay the same amount regularly.

Next infrastructure companies - which charge by use (by GPU/CPU hour used, API called, or GB stored), and amassed customer bases with spending habits that vary by the month - adopted ARR. They argue that though customers’ usage wasn’t predictable on a small scale, over a large enough customer base, customer behavior is similarly predictable to contracted spend.

Then, consumer subscription/prosumer businesses with small monthly payments began pitching using ARR. After all, there’s not much difference between an SMB paying monthly and a consumer paying monthly. 

However, there is a meaningful difference in net-dollar retention, because most consumer companies have negative NDR, and most B2B SaaS companies benefit from positive NDR (accounts that expand over time).

There’s also an ARR metric for SaaS companies called cARR, which is contracted ARR - the annualized value of contracts signed but not deployed.

And now, even marketplaces have shifted from GMV to annualizing net transactional revenue and calling it ARR…

Should all these companies benefit from the same ARR multiples? No.

And investors certainly do apply somewhat different multiples to different revenue quality.

I’m certainly getting tired of “ARR marketing” that looks made up when you look closer. Not the real deal.

We're often comparing apples and oranges.

What about you?

#artificialintelligence

Founders and AI builders in Stockholm:

This you don't want to miss this Sunday.

Apply to join here: https://luma.com/113lmm1q

CC: Anton Osika & Max Junestrand 

#artificialintelligence

Europe take note:

China is dominating open weight AI.

In less than 6 months, cumulative downloads of Chinese models ha not only overtaken US models, but began to open a widening lead. 

A good chunk of that Chinese overtake is driven by Qwen’s popularity as an open-source model - at the expense of Meta's Llama. 

Mistral has unfortunately never really been in the game, but there's time. We should be inspired by China.

The race for model dominance has only begun, and the leaderboard will look different in a year. 

Open weight is catching up to proprietary models.

Change happens gradually, then suddenly.

#artificialintelligence

When someone says "Cursor is just a $30B wrapper over VS Code and Claude" show them this 👇

It's a high-level architecture of an open source alternative. 

Cursor probably has the same + hundreds and hundreds of other user-induced features.

E.g. Cursor's tab completion model is legitimately fantastic and for many people is worth the entire $20 subscription. Lint fixes or syntax-level refactors are guessed and executed instantly with Tsb with close to 100% accuracy.

Cursor recently trained a new a new Tab model with RL that is now the default in Cursor.

This model makes 21% fewer suggestions than the previous model while having a 28% higher accept rate for the suggestions it makes.

Their Tab model runs on every user action, handling over 400 million requests per day. As a result, they have a lot of data about which suggestions users accept and reject. They use this data to improve Tab using online reinforcement learning.

We need to stop the whole "wrapper" discussion by now.

Plus, distribution is everything in the age of AI, as we've seen especially in code gen. 

In markets where the underlying technology improves faster than you can build, owning distribution and mindshare even beats technical excellence (alpha trumps levered beta, where beta means riding market trends). But it's still core to have technical excellence and shipping velocity.

Examples like Cursor show that the AI market is a motion market, not necessarily a moat one.

For a while, Microsoft’s Copilot sucked the oxygen out of the dev tools conversation. Many hesitated to build, thinking Microsoft had already won. 

Turns out, that was just a temporary illusion.

The truth is, most incumbents are still chasing AI the way they once chased mobile - late, reactive, and often allergic to taste. They struggle to design from first principles, move slowly, are bloated, and chase OKRs.

We’re still early in the AI application layer, and there has never been a better time to be a technical builder.

The race for tech markets has traditionally been framed as:

“Incumbents have distribution, while startups have tech and innovation.”

Or one of my favorite quotes:

“The battle between every startup and the incumbent comes down to whether the startup gets distribution before the incumbent gets innovation.”

But in the age of AI, it’s more like:

“Foundation models have the tech; startups have the product that gives them distribution.”

#artificialintelligence

Another of many examples why I'm so bullish about the New Nordics 👇

AI product analytics @ userjourneys.ai

sometimes I explain to people that Estonia is the unicorn and model capital of the EU, but recently I discovered that my tiny country is also objectively smarter

turns out we also have the most IOI (International Informatics Olympiad) medals per capita.

i was at a nordic dinner with Oliver Molander in sf and I told him that Estonia has ~75 IOI medals.

he said “No way”.

i looked it up, Estonia has 68 medals. that’s insane.

context:
🇺🇸 USA 127 medals (300x more people)
🇮🇳 India 70 medals (1000x more people)

no wonder recruiting insanely talented engineers in Estonia feels easy.

taplio

Oliver Molander's Linkedin Analytics

Oliver Molander

Performance ScoreOliver Molander's LinkedIn Performance

What is Oliver talking about?

Who is engaging with Oliver

Oliver Molander's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Oliver Molander's Linkedin Analytics

Oliver Molander

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Oliver Molander's LinkedIn Performance

What is Oliver talking about?

Who is engaging with Oliver

Oliver Molander's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Oliver Molander's Linkedin Analytics

Performance ScoreOliver Molander's LinkedIn Performance

Oliver Molander's Best Posts (last 30 days)