Nabeel Hyatt, GP @ Spark Capital: To Win in AI, Investors Need to Change Their Approach | E1255
Summary
AI is pushing venture from spreadsheet-solvable puzzles back toward mysteries requiring first-principles judgment. Nabeel Hyatt argues that no one knows what a model will do next week, so firms optimized for SaaS metrics, coverage, and consensus risk becoming obsolete. The host worries that spreadsheet-driven investing could make firms “dinosaurs.” Winning firms will be small, curious, and comfortable making subjective bets amid fog of war.
Venture’s organizational incentives increasingly reward markups rather than enduring outcomes. A principal “is not actually waiting for an exit. They just want a promotion, man,” so the rational move is learning what the next-stage fund wants, investing one month earlier, and packaging a four-month markup. That system is especially fragile when nobody can predict what will be hot nine months later.
Revenue velocity is no longer a sufficient proxy for AI-company quality. Products can reach $10 million ARR within months and still “probably be dead in 2 years”; the investable question is whether the founders possess enduring taste, exceptional execution speed, and the capacity to reinvent themselves. “This industry is all about exceptions,” making checklist investing structurally mismatched to the asset class.
Product matters chiefly as evidence of how founders think. Hyatt does not infer investability from a polished interface; he interrogates the decisions embodied in “the thing that comes out of their hands” to distinguish real executors from hucksters. The strongest AI founders combine speed with the taste to discard acceptable metrics and rebuild when a product “doesn’t feel right.”
Price is a test of conviction, but excess capital can alter—and kill—the company being underwritten. Hyatt is not a value investor, yet a $25 million round may create a fundamentally worse business than a $10 million round by changing hiring and execution before the company is ready. Deployment-focused investors therefore cannot assume the same eventual $5 billion outcome after forcing in more cash: “That company will probably raise another hundred million and then they might just die.”
The attractive AI opportunities sit beyond adaptation, in evolution and especially revolution. Hyatt separates incumbents adding AI, products creating genuinely new workflows, and businesses possible only because the technology exists; Spark avoids mere adaptation and seeks behaviors that “sear into your brain.” Much of today’s incubator output is instead a thin arbitrage engineered to show roughly 10% weekly growth before a seed round.
Anthropic’s bull case is a full-stack learning loop, not an indefinitely protected model moat. Owning the customer interface creates usage insight and “data exhaust” that can improve both product and model; Hyatt sees this as the “wisdom of experts,” exemplified by Descript learning from every expert edit. DeepSeek does not overturn his thesis because Spark never believed capital expenditure alone was the barrier to model competition.
Vertical agents become durable only when they attack hard, decade-long problems and have second, third, and fourth acts. Labor replacement can enlarge previously uninteresting niches, but an agent that perfectly solves today’s workflow may soon face 25 clones and drive the market toward zero. Hyatt’s filter is blunt: “Pick a job that’s hard”—ideally one where a hallucinating MVP earns initial demand while leaving years of innovation ahead.
Deep dive
1. AI has restored venture’s original fog of war
Hyatt borrows Greg Treverton’s distinction between puzzles, which raw analytical horsepower can solve, and mysteries, which require a journey through uncertainty. B2B SaaS became a puzzle—metrics, dashboards, playbooks, and armies of associates—while AI is unmistakably a mystery: “No one has any idea what a model is even going to do in a week.”
The host’s pushback—worth keeping—is that DeepSeek can overturn the world and another surprise may arrive three months later, making mysteries almost impossible to underwrite. Hyatt’s answer is that early venture originally worked this way: it was an “artisanal business,” not an industrial process, and firms must relearn how to navigate without calculating the destination in advance.
That requires “VC market fit” as surely as founders require founder-market fit. Hyatt wants small teams making subjective bets, using AI products themselves, and serving as curious sounding boards on unknowable questions such as multimodal user experience—not board members recycling Twitter commentary and portfolio markups.
Incumbent firms cannot adapt quickly because LPs reward stability while the market demands upheaval. Changing half or three-quarters of a team may be strategically correct yet damage fundraising before the transition proves itself three years later; the leaders deciding are often those promoted on yesterday’s SaaS markups.
2. Promotions and markups have replaced exits as venture’s clock
Hyatt traces simplistic heuristics back to partnership expansion: a room of seven partners became 25, 30, or even 500 people. “The industry today is run basically by principals, associates, and junior GPs,” whose career horizons are dramatically shorter than the time required for cash returns.
The incentive chain is corrosively clear: a principal wants promotion within two years, therefore needs markups, therefore learns what Coatue or another downstream investor wants this month, buys it one month earlier, and seeks a markup four months later. The work becomes packaging for the next buyer rather than discovering a future alongside founders.
The host invokes Jason Lemkin’s description of venture as packaging—acquire a company, wrap it, ship it onward. Hyatt “hate[s] that analogy deeply”: it is poor service to founders and a losing strategy when being right requires depth and nobody knows what will be fashionable nine months later.
More startups do not rescue the model. Adding nine principals against a 100-fold increase in companies merely encourages pattern matching and the “Brita filter version of investing”: maximize inbound, reject rapidly, and transact through the funnel. Hyatt instead starts from the fact that “this industry is all about exceptions.”
3. Small partnerships can carry large funds if judgment stays personal
Spark’s Web 2.0 and mobile-era DNA was formed when one company could look like a fart app in the morning and Uber in the afternoon, with no settled metrics. Hyatt admits that DNA left Spark “really badly shaped” for industrialized B2B SaaS in 2021, but its refusal to quadruple the team now makes it better suited to AI.
At the time Hyatt describes, Spark stayed at seven people with a six-person partnership: all write checks and work with founders. Its early-stage fund is a little over $700 million and its growth fund about twice that; Hyatt’s prescription is therefore not necessarily small checks, but small decision-making groups practicing subjective, first-principles investing.
Organizational politics grow because every measurement before realized, pre-tax capital is returned is a “false prophet.” Hyatt wants colleagues who begin with mutual respect and treat partnership debate as “a room that’s a search for truth,” captured internally by the phrase “being your brother’s keeper.”
Mentorship follows the same logic. Venture may be called an apprenticeship, but Hyatt sees self-actualization: a new partner cannot become a miniature version of Bijan or another mentor. Partners must uncover each person’s superpowers, recurring blind spots, and the founders with whom they can genuinely “fall in love.”
4. Early-stage and growth investing are different sports
COVID produced Hyatt’s deepest professional doubt because venture collapsed into one-hour Zoom calls followed by term sheets eight hours later. He considered leaving and wrote only about two checks in a year and a half; Spark’s early-stage quality declined before recovering, while its separate growth team navigated the period well.
Growth can support more hierarchy, principals, associates, numerical diligence, and calls to 25 customers. At seed, “there aren’t 25 customers,” so the investor exercises a different muscle. Hyatt’s question is pragmatic: when performing either job exceptionally is already hard, “why would I try and play two sports?”
The host argues that later-stage experience teaches what future investors and public markets will reward. Hyatt rejects the 2021-era premise: nobody knows what public markets will want from AI seven years hence, and internalizing this month’s preferences is likely to produce the wrong early-stage calls.
5. High-service venture cannot be high-volume venture
Against Keith Rabois’s view that the best founders do not need VC help, Hyatt draws on raising eight venture rounds as a founder. Most investors were “fine,” but an executive, assistant, or board member who is merely fine is not a success; accepting that standard means institutionalizing mediocrity.
Useful service requires emotional investment, product use, and enough detail to understand how co-founders fight and where an executive team is breaking. If no excellent partner is available, Hyatt accepts the harmless, high-priced “no-op VC”; he objects only when “do no harm” becomes the aspiration.
The model has an explicit capacity constraint: Hyatt makes roughly two to four investments annually. “You can’t have high service and high volume. Absolutely not.” His own formulation is simpler still: “I need to do a good deal a year.”
Loyalty does not exclude tough love. When things have not gone well, Hyatt repeatedly finds founders who were conflict-avoidant—a trait difficult to detect during courtship—or founders whose speed and taste were mismatched to the opportunity. Every company has “a thousand problems,” and avoiding them compounds the damage.
6. Founder quality lives at the intersection of speed and taste
Execution speed and judgment pull in opposite directions. A reflexive “shoot first, ask questions later” founder may chase yesterday’s shiny object; a person with exquisite taste may never ship. As AI automates more execution, Hyatt thinks taste matters increasingly, but the opportunity determines the correct position on the spectrum.
In a field of 25 competitors with a visible roadmap, the founder must be among the planet’s fastest executors—able to observe and aggregate everyone else’s innovation. A new-market company needs more introspection because running faster toward a derivative product does not create the market.
Granola is Hyatt’s specimen of taste under pressure. Between seed and the Spark-led Series A, Chris completely reset a product whose internal metrics were acceptable because “it didn’t feel right to him.” The achievement was not taste instead of speed, but knowing which muscle the moment required.
The host worries that OpenAI or another foundation-model provider can erase an application overnight. Hyatt concedes that static barriers offer little comfort: founders must continually reinvent. This is not an era where an eBay-like product remains essentially unchanged for 40 years; it is “a sea of speed and taste at the same time.”
7. Product reveals the people behind the pitch
Hyatt resists being called a product investor in the superficial sense. Product is “an instantiation of what the founder does”—observable evidence from the petri dish of people behind it—and therefore a way to distinguish a compelling presenter from a genuine executor.
The revealing questions concern decisions, not TAM slides: which product choice makes the candidate proud, what would they build without current constraints, and what are they embarrassed to have built? The answers matter less than the reasoning, because the product is where founders have spent their deepest hours.
His meeting cadence follows that search. A 30-minute first conversation tests chemistry; if it exists, Hyatt prefers jumping to a two-hour walk. He would probably not invest without meeting in person and often relies on long relationships—he knew Jason from Discord for seven years, despite having passed twice on Discord.
Speed need not eliminate intimacy. Wordware’s round was highly competitive and carried richer term sheets, yet Hyatt had dinner, took morning walks, and met the founders six or seven times within one week: “You don’t have time. But in the span of a week—because you care.”
8. Excess capital changes the asset being priced
Hyatt is “not that price sensitive” and rejects simplistic claims that either hot or overlooked deals systematically win; venture is made of exceptions. Yet every investment has a number beyond which economics fail, and the host exposes the slippery ladder from $60 million to $80 million to $100 million.
Hyatt calls valuation a test of conviction: liking a company at $60 million but not $65 million is incoherent. In Spark’s concentrated model, however, the deeper question is often check size and ownership—whether the business should absorb $5 million, $10 million, $15 million, or $20 million at that moment.
“Too much capital can mess up a company.” A $25 million round may kill a business that could thrive on $10 million, invalidating the growth investor’s assumption that paying up merely reduces the multiple on the same future $5 billion outcome. The extra cash may instead increase the probability of eventual death.
Figma is Hyatt’s clearest regret: Spark considered writing a very large pre-launch check, and he still thinks the missed partnership with Dylan would have been a rewarding five-to-ten-year journey. Even so, Hyatt will not divert much attention into secondaries; understanding the future remains the primary job.
9. New behavior matters more than inherited market size
Spark largely ignores market-size calculations unless they confirm an obviously narrow opportunity. Hyatt distinguishes “good businesses” from Spark’s chosen hunting ground: the firm does not need to canvas every possible winner through a Brita filter.
The signal for a new market is a new behavior that, once experienced, “just sears into your brain—you can’t stop thinking about it.” Hyatt describes himself as “a kid in a candy store” because AI produces more of these openings, though truly 10-times-better experiences remain rare.
His mobile-era lens divides AI companies into adaptation, evolution, and revolution. Adaptation adds AI to the incumbent form; evolution creates a medium-native workflow, as Instagram did versus Flickr and as Granola, Replit Agents, and Descript do today; revolution creates a platform that could not previously exist, with Uber as the canonical mobile example.
Most current startups are adaptations wearing AI paint or mediocre evolutions manufactured for an incubator deadline. Spark wants no adaptation exposure and leans toward higher-risk revolution, with selective evolution, because those are the journeys capable of producing both large exits and satisfying work.
10. Anthropic’s advantage is a full-stack learning loop
Hyatt’s Anthropic bull case begins when model builders exhaust readily available data. Improving then requires understanding what users want, and a widely used interface produces both direct insight and behavioral data exhaust unavailable to an isolated academic lab—even if that lab creates a faster algorithm.
The host’s challenge is pointed: OpenAI may have roughly 10 times Anthropic’s consumer data, DeepSeek reached number one, and products such as You.com offer credible interfaces. Hyatt concedes interface alone is insufficient; scale and growth matter, but owning the customer relationship gives a company the chance to iterate ahead.
Anthropic’s Artifacts is his example of customer insight becoming product innovation, subsequently copied by OpenAI. Model quality is only one component alongside taste, execution speed, direct customer contact, and vertical integration from the model to the user’s intended outcome.
DeepSeek therefore does not materially change Hyatt’s strategy. Spark could not have invested in Anthropic had it believed Sam Altman’s capital advantage made competition impossible; it never treated $100 billion, $500 billion, or $1 trillion of training spend as the sole moat. Hyatt’s newer rule is broader: do not evaluate any company—even a model company—by its underlying models.
11. Data exhaust compounds only when the users are exceptional
Descript observes every edit turning rough footage into a polished production, thereby internalizing expert judgment. Hyatt contrasts Web 2.0’s “wisdom of crowds” with AI’s emerging “wisdom of experts”: the objective is not an average human answer, but what the best practitioner in each field would do.
The promise is that “this little alien in your computer” helps users approach that expert standard. A top-of-market product attracts the best practitioners, whose actions can improve both the interface and the models embedded beneath it.
The host sees vertical agents expanding small markets by replacing labor, citing HappyRobot’s automation of brokers calling truckers. Hyatt agrees the opportunity can be large but warns that many such businesses are near-term arbitrages: 25 competitors may deploy similar calling agents and push the economic value toward zero.
His alternative is to “pick a job that’s hard.” The host objects that forecasting model capabilities six or nine months out is nearly impossible; Hyatt reframes the task as choosing a problem worth a decade, where today’s hallucinating MVP earns a little traction and leaves room for second, third, and fourth acts.
12. Founders should optimize for surprise, not the next-round checklist
When founders ask what metrics will secure the next financing, Hyatt redirects them from VC expectations toward the future they want to invent. If everyone already expects $8 million to $10 million ARR, merely delivering it may not attract investment: “What they want is for you to exceed expectations.”
Immediately after investing, he proposes a one-page table of contents for the pitch 18 months ahead—story, product, data, or something else—and works backward. Only about 20% to 25% of founders embrace the exercise; “people default back to numbers when they have no other story to tell.”
On enterprise companies at $8 million to $20 million ARR still doubling but judged against explosive AI revenue, Hyatt offers an unusually honest answer: “I don’t know.” Even calling them stagnant is unfair, yet he cannot say how a growth market recalibrated by Lovable and similar companies will treat them.
The broader craft remains creative rather than fixed, closer to a musician confronting a second album than an athlete improving within stable rules. Hyatt studies successes more than losses, argues for a “nothing-to-lose” mentality in investing, remains bullish on America’s long term, and regards transactionality—whether in packaging founders or debating preferred versus common shares—as “the enemy of what I’m trying to work on.”