Vince Hankes on Thrive’s OpenAI Investment and AI Competition

Vince Hankes says two questions matter for a company: who is the customer, and what is the product? The Thrive Capital partner discusses his investing career and AI with Harry Stebbings. Thrive invested in OpenAI, a material interest when assessing his views on the company and its competitors.

Hankes studied business and accounting, joined Goldman Sachs and worked on large transactions before encountering Tiger Global through work on Flipkart. At Tiger he invested in private companies alongside Lee Fixel, then joined Thrive after meeting the team while following Airtable.

He contrasts Tiger’s financial rigor and hedge-fund roots with Thrive founder Josh Kushner’s product and founder-oriented perspective. Both seek leading companies, but Thrive may ask how a salesperson would qualify a prospect rather than only asking for an ideal customer profile, in order to understand the founder’s operating problem.

Asked what motivates him, Hankes says his experiences and relationships have shaped him, but he is still learning about himself. He aims to keep a steep learning curve and increasingly sees colleagues, friends and family as people with whom learning can compound.

He says trust with founders grows from predictability: explain how decisions will be made, work through problems together and avoid surprises. Thrive’s small team, he says, gives colleagues considerable trust and autonomy.

Remaining level-headed is learned through volatility, Hankes says. In good times investors can over-extrapolate and in bad times under-extrapolate. He tries to separate emotion from the core decision while recognizing that intuition also has a place.

Stebbings asks whether AI is a lasting platform shift or another hype cycle. Hankes says it is hard to time speculation, and contrasts crypto’s ideological appeal and user-experience trade-offs with AI products already offering practical value. His judgment reflects a venture investor’s perspective.

He recalls that even major technology companies were expensive at the dot-com peak and says an investor in Microsoft at that point had to wait many years to break even. The example is his historical illustration, with no detailed return calculation supplied. His response is to examine customer value and specific product capabilities rather than chase the cycle alone.

He points to search, workflow automation, content production and learning as potential applications. Boards and founders, he says, must consider how a technology shift could disrupt or improve their businesses over several years; those impacts remain uncertain.

Thrive began following OpenAI after seeing startups build marketing and writing products on top of its models. Hankes recalls a private demonstration of work including GPT-4, after which he and Kushner saw a potentially discontinuous change in software. They then studied the product, customers and finances.

He says the investment was not obvious to everyone and that some investors declined. In a very early technology cycle, he argues, a precise market size and moat are difficult to establish; he weighed possibilities such as disrupting search against the many unanswered questions. These are the investor’s reasons, not verified future outcomes.

Rather than rely on a precise total addressable market, Thrive considered how earlier platforms scaled, citing the iPhone, AWS and Google. Hankes acknowledges these are exceptional examples. He believes AI may be comparably transformative, but that is a forecast by an OpenAI investor.

Stebbings characterizes OpenAI as a bet on one model ruling the market. Hankes says that is not his characterization and explicitly says he is speaking for himself, not the company. He points to OpenAI’s open contributions including CLIP, Whisper and Triton, while noting the core GPT model was offered through an API behind a paywall.

He argues model choice involves reliability, infrastructure, scalability and cost as well as raw capability. Some firms will use open models, but he expects many resource-constrained startups to prefer a managed service. He cites a large price reduction with GPT-3.5 Turbo as an example of changing costs; that is a time-specific comparison.

On commoditization, Hankes predicts the ecosystem around a model—distribution, plugins, multimodal interfaces and developer convenience—could be more durable than raw output alone. He does not dismiss model quality, but expects customers to value the whole platform. Thrive’s investment gives him a financial interest in this thesis.

He identifies large technology companies as serious competitors, including Microsoft despite its OpenAI partnership. He says AI talent and distribution are concentrated at large firms, and notes how quickly Microsoft, Adobe and others were shipping AI features. A startup cannot assume incumbents will remain slow.

Stebbings suggests OpenAI has startup speed and incumbent-like distribution. Asked whether startups or incumbents will capture most value, Hankes says it is too early to know. Competing directly on an incumbent’s home turf is difficult; creating a genuinely new category may favor a startup.

Hankes expects substantial value to accrue to applications because that is where customers experience it, while infrastructure providers can benefit as platforms or toll roads. His comparison to cloud software is a hypothesis about future value allocation, not a measured AI-market result.

Investing in a fast-changing market means backing extraordinary founders able to experiment and adapt, he says, but not investing blindly. If a team cannot explain the customer and product clearly, or its value might disappear almost immediately, he would wait for more clarity.

On regulation, Hankes says he is not speaking for OpenAI. He expects regulation to be necessary and wants informed dialogue between regulators and builders, warning against poorly designed rules. He describes OpenAI’s safety work positively while acknowledging he does not know everything done internally; Thrive is an investor.

Stebbings worries regulators lack the technical knowledge to keep up. Hankes considers the gap solvable through education and consultation, though he expects solutions may take years and says he does not personally know the relevant regulators. This is a policy preference, not evidence that a specific approach will work.

Stebbings cites a reported $29 billion price for the OpenAI round and asks about upside. Hankes does not independently confirm that number in the exchange; he says the price was high and the team debated it intensely. A good return at that level would require an exceptional outcome.

He says conventional models struggle to represent the potential pace of an unusual technology’s adoption. Thrive weighed financial rigor against uncertain upside in areas such as search or workflow automation. He mentions early ChatGPT Pro traction while explicitly questioning whether it would be a durable revenue engine.

A financial model is one decision tool, he says, and its weight should depend on context. Some companies have looked expensive early and cheap in hindsight, but that does not make any particular valuation sound. Thrive chose to invest based on its assessment of potential and risk, not on a proven return.

Writing a large check can be intimidating. Hankes credits Thrive’s team culture with allowing debate and collective responsibility rather than blaming one person if an investment fails. He says the small team invests across stages and tries to maintain the psychological safety to make consequential decisions.

He attributes that culture to trust, vigorous disagreement in investment discussions and warmth afterward. Stebbings asks how Kushner creates it; Hankes says leadership models both direct debate and empathy, while acknowledging that such a culture is easier to describe than build.

Hankes says his investing has evolved from a predominantly financial lens toward deeper empathy for the customer and the work of building a company. A sales-intensive product, middleware product and creative product may require different founders and organizational structures.

The strongest companies, in his view, can explain simply who their customer is and what their product does, and that clarity reaches managers throughout the organization. It helps teams make decisions consistently even when it is hard to quantify on a spreadsheet.

His memorable miss was Canva while at Tiger. He says he applied enterprise-software patterns too narrowly, focusing on churn and engagement relative to deep enterprise workflows, and missed what made the product distinctive. He now tries to ask whether the metrics explain and reinforce a company’s particular strengths rather than force a generic template.

Asked for a biggest win, Hankes declines to name one and says investors still have much to prove. He emphasizes authentic founder relationships: each investor brings a different style, and trust requires time rather than a transactional pitch.

In quick fire, he describes reading across formats and tracing the history of technologies. He declines to choose another multistage firm, and names Nat Friedman’s investing as interesting. He also says investors should distinguish a company’s execution from favorable market momentum in good times.

He points to crypto speculation during the low-rate period as an example of hype and says some apparent successes proved fraudulent. His remarks are a broad historical judgment; he does not identify the company in the transcript.

Hankes praises investor Eric Vishria for combining operating insight, commercial judgment and effective communication. He says Kushner also demonstrates that career success need not require neglecting family and friendships, a lesson that influenced his own priorities.

Looking to 2028, he hopes Thrive will back transformative companies while building its team and preserving its culture. This is an aspiration rather than a statement of realized investment returns. Stebbings closes the interview.