Key Takeaways

  • Bessemer Venture Partners invested in Anthropic's Series D at a valuation of $15 billion pre-money and $18 billion post-money.
  • Partner Byron Deeter told Sameer Dholakia that Anthropic was the lowest-voted deal in his history at the firm, proving that top venture returns require backing unconsensus convictions.
  • Bessemer passed on OpenAI at a $30 billion valuation, then watched its run-rate revenue surge from $40 million to north of $1 billion in record time.
  • Dholakia underwrote Anthropic through a cloud hyperscaler comparison, betting that frontier model development will narrow to a handful of heavily capitalized winners partnered with companies like AWS.

Missing OpenAI Forced Bessemer to Rethink Scale

Passing on OpenAI at a $30 billion valuation taught Bessemer an expensive lesson about the speed of software adoption. At the time, thirty billion looked absurd for an unproven corporate structure. Then reality struck.

Dholakia watched OpenAI jump from an estimated $40 million in revenue to over a billion, and eventually closer to two billion. In his words, “Clearly we got it wrong.” The software industry had never seen a company add that much recurring revenue that quickly. It proved that demand for raw intelligence was almost bottomless.

When Anthropic returned to the market for its Series D round, Dholakia refused to repeat the error. He evaluated the company at a $15 billion pre-money and $18 billion post-money valuation. While many growth investors recoiled at buying minority equity at an $18 billion post, Dholakia looked at the sheer slope of customer demand across the sector. Missing a generational shift because of traditional entry-multiple discipline turned out to be far riskier than overpaying for a winner.

The Cloud Redux Thesis Won Over an Unconvinced Partnership

Getting the Anthropic deal approved inside Bessemer was an uphill battle. Byron Deeter told Dholakia that the investment received the lowest vote total of any deal during Deeter's career at the partnership. In a consensus-driven firm, that score would kill a memo. Bessemer's structure, however, allows high-conviction partners to pull the trigger on deals their peers dislike.

Dholakia framed Anthropic not as an application layer bet, but as core infrastructure: “This is going to be a cloud redux was my analogy. In the cloud era, you had a handful of hyperscalers. You have AWS and Azure and GCP.”

Frontier models require billions in compute, clusters of specialized silicon, and research teams that command seven-figure salaries. That capital barrier prevents thousands of startups from competing at the frontier. Just as cloud computing settled into three dominant providers, base model intelligence would consolidate into three or four private utilities. Anthropic secured a deep relationship with AWS, giving it the distribution and compute capacity needed to survive the capital war.

Dholakia now views his initial position as too cautious: “I wish I had the ability to convince us all to do a bigger check that round.” When the market structure points toward a concentrated oligopoly, allocating a small check to manage downside often hurts more than writing a concentrated check to own upside.

What to Do With This

Review your decision log from the past twelve months and identify the single opportunity you rejected purely on price. Recalculate that company's revenue growth trajectory against your initial baseline. If their customer acquisition outpaced your model by more than 2x, write down the structural reason your original valuation ceiling failed. Use that revised ceiling to evaluate your next supplier, hire, or technology bet this week.