Factor Capital Update - August 2025

The biggest story in tech over the past several months has been the incredible arms race for AI talent, initiated by Google and Microsoft and recently escalated to an entirely new level by Meta over the past two months. The headlines have focused on the paydays being handed out by Zuckerberg to the new Superintelligence team at Meta, led by Alexandr Wang of Scale AI, who, along with a small team of executives, was poached via a $14.3 billion payout to Scale shareholders (including Wang). The dynamics remind me of the way that college sports have evolved through Name, Image, and Likeness (NIL) money over the past couple of years, and I think there are some interesting parallels to the startup landscape, which I’ll cover in this month’s update.

The NIL deals didn't just change college sports - they ended an era. What began as a necessary correction, allowing student-athletes to profit from their talents, quickly spiraled into a financial arms race. Elite programs across the Big 10, SEC, and ACC have amassed multimillion-dollar budgets to recruit and retain top talent, creating a stark divide between the haves and have-nots. Capital has permanently reshaped collegiate competition. Previously, smaller schools could remain competitive by identifying overlooked prospects and patiently nurturing them. A talented coach with the ability to spot under-appreciated talents and develop them over a four-year career could routinely build a national championship contender.

Today, that model has vanished. Breakout underclassmen showcasing higher-level talent are around for no more than a season before being lured away by richer programs, replacing patient talent cultivation with instantaneous, high-stakes transactions. This dynamic has bifurcated college athletics, specifically in basketball and football. An elite tier of schools battles fiercely for national championships, backed by massive budgets. Meanwhile, second-tier programs must accept a narrower competitive landscape, effectively chasing different, smaller prizes. This year’s March Madness basketball tournament was a good example of this, with upsets few and far between and the later rounds dominated by the wealthy powerhouses.

The reality is clear: in the major revenue sports only teams with the deepest pockets can genuinely compete for the top prize. If you don’t have the resources to do it, then you have to focus and give up on the goal of being nationally competitive in all the department’s sports and instead just focus on being elite at one or many more niche sports where NIL plays a less meaningful role.

It is almost identical to what is unfolding in the AI industry.

Capital is becoming the primary strategic weapon, creating a stark divide in the startup ecosystem. This isn't entirely new, but rather an amplification of trends observed over a decade ago. In the late 2000s and early 2010s, the bottleneck for startups was access to engineering talent, driving salaries upward and meaning that the $500k raises that propelled Uber and Airbnb and others back then, at sub-$5 million valuations, were a relic.

Separately, the scarcity also created a safety net: top engineers at Google or Facebook could comfortably start companies, knowing that their downside was essentially that they’d likely boomerang back to big tech with engineering acquihires typically valued around $1 million per engineer waiting for them where they could just “rest and vest.”

What we’re seeing now is that AI has intensified this safety net by orders of magnitude. As discussed above, leading labs like OpenAI, Google DeepMind, and Meta now offer compensation packages ranging from eight to ten figures to secure top researchers. For startups aiming to build world-class AI teams, the acquihire floor has risen dramatically, reaching upwards of $10 million - and in extreme cases, as high as $100 million or more - per engineer, if you can package together an attractive cadre of experts for the leading labs. This places startups in an increasingly precarious position, forcing them to raise massive amounts of capital before even developing a viable product.

The capital is not merely for talent acquisition to compete in this arms race; it's crucial to sustain the immense computing costs essential for frontier AI development. This intense pressure creates an intriguing paradox. While the cost of competing at the frontier is astronomical, the outputs - foundational AI research - are rapidly commoditized and broadly accessible. Leading labs freely share their breakthroughs through open source models, granting immense leverage to businesses choosing not to compete head-on.

At HyperScience, we experienced this firsthand in 2015. We recognized that building proprietary foundational models was akin to reinventing the wheel, especially when labs like Facebook’s FAIR and Google’s Brain and DeepMind teams, as well as universities, were openly sharing revolutionary blueprints. Many companies were funded to build cutting-edge, general-purpose image recognition models, for example. None of them exist in that form today as that became commodity open-source technology.

But if you can simply apply the research put out by these leading labs in production use cases (which most businesses aren’t equipped to do), there are great businesses to be made on the back of their innovations at the edge. The most transformative example was the "transformer" architecture introduced by Google in 2017, which enabled models to grasp context in language processing more naturally. This innovation laid the foundation for OpenAI’s ChatGPT, which has not only reshaped the industry but society at large. OpenAI is now among these well-armed labs as they are in the process of raising $40 billion from private investors right now and are on track for $20 billion in ARR.

This landscape presents two strategic paths for founders in AI specifically, but I think it applies across more than just these specialized AI companies.

• The Arms Race: Raise significant capital to directly compete with industry giants for talent and compute, chasing immense equity outcomes but risking swift collapse if venture capital appetite fades.

• The Specialist’s Path: Leverage increasingly commoditized, cutting-edge AI to solve specific, targeted business problems through an AI native approach, focusing on narrower but highly lucrative markets rather than general-purpose applications.

The buzziest AI version of this is a company called Harvey, initially focused on legal work and document discovery. But we’ve seen this evolution occur over the past several months in crypto as well, where a small handful of well-capitalized and highly performant general-purpose blockchains make starting a new one from scratch seem unnecessary.

In response, we’ve seen specialized chains like HyperLiquid for trading and Story Protocol for intellectual property demonstrate the opportunity and potential reward of targeted innovation. EigenLayer, initially a broad and buzzy infrastructure layer, recently executed this type of pivot as well, focusing specifically on verifiable compute.

Some of the next decade’s greatest opportunities might lie on this specialist path. Most of the Factor portfolio fits this mold, and I suspect the companies in Fund II will continue this trend. The capital pursuing frontier advancements will continually drive down the cost and accessibility of powerful AI tools and blockchain infrastructure - you can just bank on that trend the same way you could with Moore’s Law for transistors.

And founders who focus on playing games where others aren’t as focused will have a greater chance at success and can run a different playbook as well. The new safe landing might shift away from the acquihire path and instead be one’s ability to affordably build profitable and sustainable businesses you can live very comfortably off of for many years.

This approach will increasingly attract exceptional talent beyond the limited pool of elite researchers commanding nine-figure packages. For many, owning significant equity in a profitable, specialized business will be vastly preferable to perpetually racing against industry giants. And I am betting that a different type of funding model will need to exist to support this alternative path.

Thanks, as always, for reading.

-Jake Dwyer
Founder
Factor Capital