Nabeel Hyatt on Venture Investing and the Uncertainty of AI

Spark Capital general partner Nabeel Hyatt joins Harry Stebbings to discuss venture investing, founders and AI. His assessments of companies such as Anthropic and Granola reflect Spark’s investment interests and the uncertainty of the market at the time of the interview.

Hyatt says investors must adapt their approach as AI changes the market. He contrasts a “puzzle,” which can be solved with enough known information, with a “mystery,” where the situation changes and the answer cannot be calculated in advance.

He argues that the previous SaaS cycle encouraged standardized metrics and large sourcing teams. With models and user behavior changing quickly, he believes investors need curiosity, judgment and the ability to work with founders on questions that have no settled answer.

Hyatt expects many firms to struggle to change because their organization and limited-partner relationships were built for an earlier environment. This is his competitive assessment, not evidence that any particular firm has already failed.

He favors small teams making subjective, carefully considered bets and acting as a sounding board for founders. In his view, investors should use the products and understand emerging workflows rather than rely solely on market chatter and valuation marks.

He criticizes incentives that reward junior investors for rapid valuation markups and promotion before an exit. His description is an argument about parts of the industry, not a verified characterization of every investor or firm.

Stebbings compares venture to packaging companies for the next round. Hyatt understands the analogy but dislikes it, arguing that following what later-stage investors currently want can be a weak substitute for a durable understanding of the future.

A larger team cannot automatically cover an exploding number of startups, he says. His alternative to a high-volume filtering funnel is to focus on a narrower set of opportunities and spend more time understanding founders and products.

Hyatt says Spark’s relatively small partnership grew out of earlier periods of technological change and did not expand as aggressively during the 2021 SaaS boom. He believes that structure is better suited to the present environment, an assessment tied to his own firm.

The guests discuss losing deals to competitors willing to offer much higher prices and founder-friendly terms. Hyatt says even highly priced deals can work, but resists a rule that hot deals are always good or always bad.

He asks whether a company has a genuine competitive advantage rather than merely being one of many similar providers. The answer depends on the market and the company; simple comparisons cannot establish lasting value.

Hyatt discusses partnership and mentorship as a way to help investors develop judgment. Constructive questions and shared experience can be more useful than a single prescribed playbook when a market is changing.

He reflects on doubting his own approach during the 2021 investment boom, when other firms moved faster and prices rose. His account illustrates the pressure to follow a prevailing strategy even when it does not fit one’s convictions.

On investor service, Hyatt argues that founders need different kinds of help at different moments. An investor can listen, challenge assumptions and offer context without taking over the founder’s responsibility for the company.

When confidence in a founder weakens, he says the investor must understand the specific issue and the job the company now requires. The discussion emphasizes honest dialogue and the difficulty of separating a temporary challenge from a more fundamental mismatch.

Hyatt distinguishes product taste from execution speed. Different opportunities require different combinations, and a founder who was well matched to one stage may face a changed market that demands new capabilities.

He cites Granola as an example of a founder changing the product substantially despite acceptable early metrics. Spark led a financing in the company, so Hyatt’s favorable account comes with an investment interest.

In a fast-changing AI market, Hyatt believes durable value requires continuous innovation rather than assuming a single product or barrier will last indefinitely. That is his investment thesis, not a guarantee that constant change produces defensibility.

He is skeptical of treating a market-size slide as a precise answer when a startup may be creating a new category. Spark still chooses where to focus; he is not claiming that every apparently small market will become large.

Hyatt uses products to learn how founders think. He asks why they made particular decisions and looks at what they built, not simply whether a polished pitch or spreadsheet matches a template.

He says he usually speaks with roughly one or two companies a day and prefers an initial short conversation followed by a longer one when there is mutual interest. Some relationships, including those with founders of Discord and Cruise, developed over many months or years.

Price matters, but he says check size, desired ownership and the amount of capital a company can use well are linked. He worries that an oversized round can change hiring and incentives in ways that damage a young company.

Hyatt identifies Figma as a missed opportunity associated with a large pre-launch financing. His regret concerns both the investment and the chance to work with the founder; the interview does not quantify an actual loss from a deal Spark did not make.

He does not actively make secondary transactions a primary focus, although he says they do happen. His reasoning is that finding and supporting early companies already consumes scarce time; he recognizes that investors eventually need liquidity.

Hyatt groups AI startups into adaptation, evolution and revolution. Adaptation adds AI to an existing product; evolution changes a workflow; revolution creates a new kind of product or platform that was not previously possible.

He cites Granola, Replit and Descript as examples of changed workflows and says Spark prefers more disruptive evolution or revolution to a superficial AI addition. These categories are a lens for discussion, not an objective ranking of each company.

Spark is an early investor in Anthropic, and Hyatt says he believes value can accrue both to model providers and applications. His positive case for Anthropic includes product execution, direct customer relationships and learning from usage; it is an interested investor’s prediction.

He says DeepSeek did not fundamentally change his view that capital alone is insufficient to win in foundation models. He expects strong teams to combine model development, product taste and customer feedback, but does not establish which company will ultimately lead.

Hyatt argues that feedback from skilled users can inform the next generation of products, using Descript’s editing workflow as an example. The usefulness of such data depends on how it is obtained and applied; the conversation does not establish access rights or outcomes for all products.

On AI agents, he worries that an easy-to-copy service built around today’s model capabilities can become crowded quickly. He encourages founders to choose difficult problems worth pursuing for years and plan for further innovation, rather than rely only on a short-lived arbitrage.

He discusses losses and uncertainty as part of investing and says judgments must be revisited when the facts change. A weak outcome or missed deal can teach an investor about both the original decision and the assumptions behind it.

Hyatt is skeptical about Europe’s current AI startup environment and argues that San Francisco offers valuable concentration of talent and knowledge. Stebbings counters that local hiring costs and competition are high; the exchange presents competing views, not a measured ranking of regions.

When founders ask what they need for the next financing, Hyatt tries to begin with what they want to build and what might surprise them over the next year. He believes a compelling story should emerge from the business rather than simply meeting a standard metric.

The guests acknowledge uncertainty about software companies growing at a healthy but slower rate than the newest AI startups. They do not predict a specific outcome for that cohort or imply that doubling revenue is stagnation.

In quick-fire questions, Hyatt says he has become less willing to evaluate an AI company solely by the model underneath it. He also reflects on parents who supported his unconventional path and tries to offer his own children guidance without dictating their choices.

He says having another possible career as a founder helped him approach venture with less fear of losing his position. He still recognizes that the firm must produce results for limited partners and support its founders.

Hyatt describes venture as a creative exercise in a changing market rather than a fixed game with a single correct answer. He encourages investors to navigate uncertainty with both conviction and doubt.

He recalls learning from a senior colleague to focus on whether he had done thoughtful work rather than trying to measure every day immediately. Enjoying the work, he says, does not remove accountability for long-term investment outcomes.

Hyatt closes by asking how venture could better support innovation and exceptional founders. He is open to evolving deal terms, but favors treating the founder and investors as people committed to a common purpose rather than as a purely transactional relationship.