Sam Altman’s Kingmaker Strategy: Flipping the Chip Deal, The $5B Seed Round Era, and Why Companies at $300M ARR Can’t IPO
Bottom Line Up Front:
OpenAI has fundamentally restructured the power dynamics in tech infrastructure by securing warrants to purchase 10% of AMD—essentially getting equity for the privilege of buying their chips. This stands in stark contrast to their Nvidia deal, where Nvidia invested equity in OpenAI for the right to sell to them. The move reveals that having users and revenue commitments matters more than profitability in today’s AI infrastructure wars. Meanwhile, the venture landscape is experiencing unprecedented price inflation at early stages, with proven founders raising at $5B+ pre-money valuations that challenge traditional return mathematics—yet these deals may still make sense given the capital-intensive, winner-take-most nature of AI markets.
Key Takeaways
1. Users and Revenue Commitments Create More Leverage Than Profitability OpenAI’s ability to extract equity from AMD while losing billions demonstrates that controlling distribution and having scaled usage matters more than bottom-line metrics in platform wars.
2. Traditional Venture Return Math Is Being Stress-Tested When seed rounds happen at $5B+ pre-money, even successful outcomes of $50-100B barely generate traditional venture returns. But in capital-intensive infrastructure markets with proven founders, this may be the new normal.
3. Kingmaking Works Best in Capital-Intensive, Winner-Take-Most Markets You can’t kingmake in every category, but in markets where infrastructure spend creates sustainable moats, early capital concentration can create decisive advantages—though rarely monopolies.
4. The IPO Bar Has Never Been Higher With median 2025 IPOs at $931M ARR, companies at $300M growing 25% face tough choices: continue building toward scale, consolidate with peers, or accept PE valuations well below recent private marks.
5. Founders Must Take Control of Their Destiny Companies in the “muddy middle” need three things: re-incentivized founding teams with equity for growth, profitability to eliminate capital dependency, and second acts (often AI-related) to reignite growth.
6. Deregulation Creates Instant Value Shifts Polymarket went from essentially illegal to $9B valuation with NYSE investing $2B, showing how regulatory shifts can transform entire business models overnight.
7. LP Liquidity Will Increase by Necessity As successful companies stay private for 15+ years instead of 8-10, secondary markets for LP positions will become increasingly important infrastructure, even if currently inefficient and frictionful.
8. Component Suppliers Still Get Squeezed—Unless They Have Monopolies Nvidia’s 50% margins represent an aberration from typical component economics (cost plus 20%) that only exists due to architectural lock-in and monopoly position—a situation OpenAI is actively working to dilute.
The AMD Deal: A Masterclass in Leverage
OpenAI’s chip partnership with AMD represents one of the most revealing power dynamics in modern tech. The company secured warrants to purchase up to 10% of AMD at essentially a penny per share—but only if they actually buy the chips and AMD’s stock price increases. The structure is telling: AMD had to give away equity for the privilege of having OpenAI as a customer, while Nvidia got to invest in OpenAI for the same privilege.
The deal mechanics reveal sophisticated negotiation. OpenAI essentially told AMD that their business would be so transformative for the chipmaker’s stock price that they deserved to capture some of that upside. AMD’s stock jumped 30%+ on the announcement, validating the thesis. The warrants only vest if OpenAI actually purchases the chips and the stock price remains elevated—aligning incentives while still transferring substantial value.
This creates an interesting hierarchy: OpenAI has less power than Nvidia (hence letting Nvidia invest in them) but more power than AMD (hence extracting equity from AMD). The common thread? OpenAI controls the users and has credible multi-hundred-billion-dollar compute spending commitments, even while losing money hand over fist. In infrastructure wars, revenue commitments from scaled users trump profitability metrics.
Rory O’Driscoll on OpenAI’s negotiating power:
“Sam Altman understands power. He has more power than AMD. So he took 10% of their company for the privilege of selling stuff to him. He has probably less power than Nvidia. So he let them get equity for the privilege of selling him chips. It speaks to the hierarchy—there’s dominance clearly been established.”
The deal also carries echoes of historical tech power shifts. Thirty years ago, Microsoft dominated PCs with DOS and Windows, Intel supplied the chips, IBM created the ecosystem but faded, and AMD played second fiddle with ~10% market share. Today’s parallel is striking: OpenAI dominates the AI platform layer, Nvidia supplies dominant chips, Microsoft enabled the ecosystem (and may fade relative to OpenAI’s power), and AMD again serves as the necessary second source. History doesn’t repeat, but it certainly rhymes.
Microsoft: The New IBM?
The comparison to IBM’s PC era mistakes grows more uncomfortable for Microsoft with each OpenAI announcement. At Developer Day, OpenAI positioned ChatGPT as the place where users should run all their applications—accessing Figma, Canvas, Spotify, and other tools natively within the ChatGPT interface. This is precisely Microsoft’s historical role: providing the operating system layer where applications run.
Microsoft does own a significant stake in OpenAI, which is better than IBM’s complete miss on Microsoft equity three decades ago. And to their credit, the Microsoft corporate development team deserves bonuses for at least securing ownership. But owning 30% of the monster you created is cold comfort when that monster is positioning itself as your replacement. The question for Microsoft: did we just fund our own disruption?
O’Driscoll on the PC era parallels:
“What you’re seeing here is the Windows Intel game beginning again. Microsoft was the software company that took control of the PC monopoly. IBM was the old school company that set them up. AMD was the little player that got dealt in to 10% market share. Fast forward to today: OpenAI dominates, Nvidia is Intel, AMD plays the same role, and Microsoft is IBM—they set this Viper in motion.”
The Nvidia angle adds another dimension. Despite appearing to compete just days after Nvidia’s $100B investment in OpenAI, the AMD deal was carefully orchestrated. Sam Altman was explicit in praising Nvidia as the number one vendor and sequenced the announcements thoughtfully. Nvidia, for its part, can afford to be gracious—they’re making incomprehensible amounts of money with 50% operating margins on a $200B+ revenue run rate. Jensen Huang understands he needs to give up some market share to avoid regulatory scrutiny and maintain customer relationships, even while keeping 90%+ of the market.
Jason Lemkin on Nvidia’s calculated generosity:
“The only person making any money in AI is Nvidia. Even Oracle isn’t making any money. OpenAI certainly isn’t making any money. Nvidia is making so much money it’s almost incomprehensible. So I think Jensen knows he’s got to give up some of it. He can minimize his price erosion and maximize his market share without creating a huge conflagration.”
The fundamental insight: component suppliers typically get squeezed on margins (cost plus 20% is standard), but Nvidia’s 50% margins reflect a true monopoly position with architectural lock-in. That’s the primary attack surface for anyone buying chips at scale, and OpenAI is systematically working to create alternatives.
Venture Pricing Has Gone Parabolic—And Maybe That’s OK
Naveen Rao, former VP of AI at Databricks, raised $1B at $5B pre-money for his new venture. The math is brutal: accounting for dilution, investors need roughly a $50-100B outcome for 10x returns. This represents a fundamental shift from venture’s historical pricing dynamics.
The question isn’t whether this is expensive—it obviously is. The question is whether it makes sense. In hard infrastructure markets, the pool of people who can credibly solve these problems is tiny. Rao has successfully built and exited two deep tech companies, including one to Intel and Databricks. He understands the problem space, knows all the key players, and has demonstrated grit and execution. When someone with that profile walks into your office for the third time, the decision gets a lot simpler.
Lemkin on how confidence changes the game:
“If you’re Naveen at Databricks, you’ve seen a hundred billion and more going up. So 5 to 100 seems plausible. Back in the day it was really hard to see north of a billion dollar outcome for a lot of these startups. And now it’s very easy if you are a Databricks alum to see a hundred billion or more because you were just there last week.”
The confidence game has also changed. Back when Lemkin and O’Driscoll started investing, seeing a path to even $1B outcomes was difficult. Today, if you’re a Databricks alum, you just watched your company grow past $100B. The mental model shifts from “can we get to a billion?” to “we’ve seen what hundred billion looks like, and 5 to 100 seems plausible.” Founders who’ve been in the building have different risk tolerances.
The structure of these deals also matters. Companies raising at these valuations typically maintain high growth rates even at scale. If you’re doing $100M at 50x revenue multiple and then $1B at 50x multiple, the pricing is consistent—provided you can sustain the growth and the market is big enough. The risk isn’t overpaying for growth; it’s hitting a market size wall where rapid deceleration makes the large-scale valuation untenable.
O’Driscoll on venture’s forgiving nature:
“Venture is the most forgiving equity business of getting the price wrong. PE, if you get the price wrong, they’re low variance assets. If you overpay by 50%, you’re toast because they’re 3x assets. The great thing about venture is it has maximum variance. But you can push a theory to destruction—if you pay for everything where you’ve got a 5x return if you do something wildly amazing, and only one in three does something wildly amazing, you don’t have much of a return.”
Venture remains the most forgiving asset class for getting pricing wrong, precisely because of its variance. In private equity, overpaying by 50% on a 3x asset destroys returns. In venture, exponential growth can overcome substantial overpayment. But this only works to a point—if you’re paying for 5x outcomes on everything and only one in five delivers, the math breaks down regardless of variance.
The Kingmaking Phenomenon: When Capital Creates Winners
The debate over “kingmaking”—whether venture capital can pick winners by flooding them with money early—revealed nuanced thinking. The simple answer is yes, kingmaking happens, but it’s complicated.
In the largest markets, kingmaking is clearly occurring. OpenAI’s capital strategy has made it extremely difficult for competitors to raise similar amounts. When you’ve locked up $100B+ in commitments, that capital isn’t available for X.AI, Safe Superintelligence, or other recent high-profile AI startups. It’s not a monopoly—Anthropic clearly competes—but it’s definitely an oligopoly that falls away quickly after the top 2-3 players.
The phenomenon extends beyond foundation models. In markets like vibe coding, legal AI, healthcare AI, and customer service AI, early leaders that secure tier-one venture backing and deploy $50-200M can create difficult competitive dynamics. Not insurmountable—Legora came from Sweden to challenge Harvey’s apparent kingmaking in legal AI—but definitely challenging.
The key variable is market size. In enormous TAMs, multiple well-funded players can coexist. In smaller markets, kingmaking becomes more decisive. But the mechanism has changed from historical software investing. Companies used to catch up around $10-20M ARR where execution mattered more than capital. In AI infrastructure, capital advantages often invert and become more valuable at scale due to compute costs, talent acquisition, and distribution requirements.
Lemkin on the changed dynamics of AI investing:
“If I believe I need a hundred million to scale to compete with Replit and Lovable, then I’m gonna bow out and sell because there’s no option. That’s what’s different than classic 80% gross margin software. In the age of AI, I’m not sure you lose the capital advantage around 20 million ARR. I think it often is inverted where that capital is more helpful.”
There’s an important caveat: calling it “kingmaking” gives venture too much credit. The entrepreneur creates the initial success—getting to $2-4M in revenue with strong unit economics and a differentiated product. Venture capital then amplifies that success through brand (attracting follow-on rounds), capital (building moats), and network effects (deterring competition). But the fundamental act of creation remains entrepreneurial.
IPO Math: The Snyk Reality Check
Snyk’s situation crystallizes the challenges facing late-stage private companies. At ~$300M ARR growing 26% (down from 150% in 2022), they’re right at the edge of IPO viability. The 2025 IPO market has been strong, but selective: the median IPO this year hit $931M in revenue run rate. Snyk isn’t miles away, but they’re just below the threshold.
This creates three options: find a PE buyer, seek a strategic acquisition, or consolidate with peers (like DBT and Fivetran did to reach scale). The problem? PE buyers aren’t aggressively circling despite attractive Rule of 40 profiles. Multiple portfolio companies in similar situations—one at Snyk’s scale, one at a third the size, one at a tenth—have received zero PE offers. The capital exists, but there’s no rush to deploy it for sub-scale assets without defensible market niches.
Lemkin on PE market reality:
“I have three companies that should have gotten PE offers. None of these have had PE offers. Zero. They all have the right Rule of 40 numbers, the right NRR. Why not buy them? They’re not going to IPO, but they’re good companies. Crickets. Crickets from the PEs.”
The solution requires taking control of destiny in three ways. First, re-incentivize founding teams with “equity for growth” grants. Many founders are fully vested after 10+ years but face another 5-7 years of building. Giving them 3-4% more equity tied to growth milestones realigns incentives without massive dilution. Second, achieve profitability to eliminate capital dependency—if you’re not reliant on external funding, market timing matters less. Third, find a second act, ideally tied to AI trends, that can reignite growth trajectories.
O’Driscoll on taking control of destiny:
“The number one thing I’d say to these companies is you got to take control of your destiny. Make sure the management and founding team is excited and have something to play for. I call them EFGs, equity for growth. You say to the founder, ‘You’re fully vested, but here’s another seven years of equity ahead of me if you deliver growth.'”
The alternative is being forced into a down round or fire sale when the liquidity window closes. And liquidity windows in private markets open rarely and unpredictably. When they open, companies must pay attention even if they ultimately decline. The lesson: asymmetric bet on founder longevity and excitement. The Aaron Levies, Drew Houstons, and Mike Cannon-Brooks who sustain 15+ year journeys are exceptional precisely because most humans aren’t wired for that duration.
Vercel at $9.3B: Following the Developers
Vercel’s $300M raise at $9.3B valuation prompted suicide round concerns, but the underlying thesis is straightforward: follow where developers are going. Vercel has become the default hosting platform for modern web applications, just as Supabase became the default for Postgres management. These are captain obvious bets on fundamental infrastructure shifts.
Lemkin on following obvious trends:
“The more you do this, the more you just say to yourself, you just need to do big exciting deals in trends that are absolutely obvious. And every time you try and make it harder than that, you just lose money.”
The “suicide round” label assumes insufficient capital for the valuation bar set. But $300M provides substantial runway, and if the company was already profitable or near-profitable, it’s not a suicide round—it’s strategic positioning. The risk comes from market size constraints. If the total addressable market for Vercel-style hosting can’t support a $50-100B outcome even with dominant share, then paying 20x+ revenue multiples at scale becomes problematic.
What makes the pricing potentially defensible: growth rates aren’t declining. If you paid 50x revenue at $100M ARR and the company hits $1B ARR still growing 50%+, paying another 50x multiple is logically consistent—provided the market keeps expanding. The phenomenon of maintaining revenue multiples across order-of-magnitude valuation increases is unusual but justifiable in winner-take-most infrastructure markets.
The broader pattern holds across AI infrastructure: proven products in obvious developer workflows can command premium pricing if they maintain growth at scale. The question is always market size and competitive sustainability, not whether the current execution justifies current pricing.
LP Liquidity and the Secondaries Market
Brown and Northwestern selling VC stakes follows Yale and Harvard’s similar moves earlier this year. The driver is twofold: university endowments face political pressure requiring more liquidity, and successful venture funds now take 15+ years to realize rather than the historical 8-10 years.
The experience of managing LP secondary sales reveals a broken market. LPs typically have zero rights to sell—literally none—forcing awkward negotiations where they claim moral authority rather than contractual rights. The process is frictionful, inefficient, and prices in significant discounts. Yet it’s becoming structurally necessary.
Better liquidity benefits both sides, especially as companies deliberately stay private longer. Supporting long-term LPs who want to clean up books after 10-12 years makes sense when one tail-end company remains. The alternative is forcing LPs to wait 15-20 years for full distributions, which doesn’t align with their own obligations and timelines.
O’Driscoll on liquidity’s real constraint:
“Liquidity doesn’t evaporate because people run out of money. Liquidity evaporates because people get scared and want to keep their money. At some point when that happens, you’ll go, ‘Oh, that’s what the public markets were for.'”
The critical insight: liquidity doesn’t evaporate because people run out of money. Liquidity evaporates because people get scared and want to hold cash. When that happens—and it will—public markets suddenly look attractive again. The long private period works wonderfully when private capital flows freely, but poorly when risk appetite contracts. This argues for faster paths to public markets, not because companies aren’t ready, but because optionality has value.
SPACs Return with Better Terms
Chamath’s new SPAC launched with notably improved terms: sponsors now only profit if the stock achieves at least 50% gains. This fixes the previous generation’s worst incentive misalignment, where sponsors made money simply by completing deals even if investors lost money.
The structure still isn’t cheap—sponsors get a 30% promote once the stock hits target thresholds—but it’s more rational than before. The regulatory arbitrage remains: SPACs can make forward-looking statements that traditional IPOs cannot, allowing “pumping” that S-1 filings prohibit. But at least dumping now requires creating actual value first.
Whether this makes SPACs competitive with well-run IPOs remains questionable. The improved terms move them from “obviously worse” to “potentially reasonable alternative for specific situations.” But for high-quality companies with strong advisors, the traditional IPO path likely still wins.
Polymarket’s $2B Round: Deregulation Creates Instant Value
The most striking deal might be Polymarket raising $2B from the New York Stock Exchange at a $9B valuation. The platform was essentially illegal in the US just months ago, facing shutdown threats from the Biden administration for offshore gambling. Trump’s election, his son joining the board, and David Sacks’ deregulation push transformed the company from pariah to darling overnight.
The Intercontinental Exchange (owner of NYSE) isn’t making a pure financial investment—they’re buying strategic position in an emerging market for prediction platforms and real-time sentiment data. The $2B represents roughly 20%+ of the company, suggesting exclusivity arrangements and data sharing beyond simple capital deployment.
Whether the business evolves beyond sports betting (currently 70-80% of volume) into legitimate prediction markets for elections, policy outcomes, and business events remains the key question. If it does, the infrastructure for electronic market-making that ICE specializes in becomes extremely valuable. If it stays primarily sports betting with political markets as marketing, the valuation becomes harder to justify.
The broader lesson: regulatory shifts create or destroy entire business models instantly. Companies in regulatory gray areas should be prepared to move quickly when windows open, as Polymarket did. The timing is everything.
Quotable Moments
Rory O’Driscoll on OpenAI’s negotiating power:
“Sam Altman understands power. He has more power than AMD. So he took 10% of their company for the privilege of selling stuff to him. He has probably less power than Nvidia. So he let them get equity for the privilege of selling him chips. It speaks to the hierarchy—there’s dominance clearly been established.”
Jason Lemkin on Nvidia’s calculated generosity:
“The only person making any money in AI is Nvidia. Even Oracle isn’t making any money. OpenAI certainly isn’t making any money. Nvidia is making so much money it’s almost incomprehensible. So I think Jensen knows he’s got to give up some of it. He can minimize his price erosion and maximize his market share without creating a huge conflagration.”
Harry Stebbings on kingmaking and market size:
“The smaller the TAM, the more prominent the kingmaker ability is. In law for example there are many who have got a lot of funding and a lot of traction. Same in healthcare, same in customer service, same in coding. But then as you go smaller markets, kingmaking really becomes more prevalent.”
Lemkin on how confidence changes the game:
“If you’re Naveen at Databricks, you’ve seen a hundred billion and more going up. So a $5 Billion “Seed” round to $100 Billion+ outcome seems plausible. Back in the day it was really hard to see north of a billion dollar outcome for a lot of these startups. And now it’s very easy if you are a Databricks alum to see a hundred billion or more because you were just there last week.”
O’Driscoll on venture’s forgiving nature:
“Venture is the most forgiving equity business of getting the price wrong. PE, if you get the price wrong, they’re low variance assets. If you overpay by 50%, you’re toast because they’re 3x assets. The great thing about venture is it has maximum variance. But you can push a theory to destruction—if you pay for everything where you’ve got a 5x return if you do something wildly amazing, and only one in three does something wildly amazing, you don’t have much of a return.”
Harry Stebbings on the suicide round question:
“I’ve heard these described before as suicide rounds—like a super high price with actually not such a huge amount going in, 300 million. So it sets a huge expectation with not massive capital injection. Does this just break venture math? These deals are unusual. A billion at 5 billion for me to get a 10x it needs to be a hundred billion company.”
Lemkin on the changed dynamics of AI investing:
“If I believe I need a hundred million to scale to compete with Replit and Lovable, then I’m gonna bow out and sell because there’s no option. That’s what’s different than classic 80% gross margin software. In the age of AI, I’m not sure you lose the capital advantage around $20 million ARR. I think it often is inverted where that capital is more helpful.”
Lemkin on PE market reality:
“I have three companies that should have gotten PE offers. None of these have had PE offers. Zero. They all have the right Rule of 40 numbers, the right NRR. Why not buy them? They’re not going to IPO, but they’re good companies. Crickets. Crickets from the PEs.”
O’Driscoll on founder endurance:
“The astonishing thing about your Aaron Levies or your Drew Houstons or your Mike Cannon-Brooks is the longevity. It’s just so few that actually really do the 15-year journey. Most humans should take the millions and relax. They’re not wired this way.”
