Is AI the new cigarettes? Our VC panel digs into Anthropic’s S-1 | E2356

Anthropic’s proposed Nasdaq listing under ticker ANT left its price range and share count blank, while a report cited on the show said the company aimed to go public the week of November 16, seeking at least $100 billion at a valuation of roughly $1.5 trillion to $2 trillion. The panel saw the S-1 as an attempt to reassure investors about profitability and spending, but disagreed on how much it revealed. Dave McClure was bullish at $1.5 trillion, but not at $2 trillion to $2.5 trillion; D.A. Wallach said the disclosures were too fragmented to establish the underlying picture.

Anthropic reported $16.3 billion in first-half 2026 revenue, 14 times the roughly $1.2 billion reported in the first half of 2025, and $974 million in Q2 operating profit. Its $80.7 billion first-half net loss included $72.4 billion in non-cash fair-value charges, mostly tied to convertible-note agreements with Amazon and Google. The company remeasures those liabilities each reporting period, treating increases as expenses and adding them back in cash-flow reporting. McClure said excluding the charge left Anthropic in good shape on profitability and cash outflows, while cautioning that Amazon-related notes may have subsidized AWS compute and obscured true costs.

The panel questioned the quality and durability of the growth signals. McClure said growth appeared to have slowed, though he did not think Anthropic had stopped growing; he cited possible share gains by OpenAI or open-weight models, as well as flawed data. A chart showed a flat $76.8 billion annualized revenue run rate, but its source was unclear, and he stressed that a run rate is not recurring revenue. Another chart put Anthropic at a $65 billion run rate, versus $50 billion for OpenAI. McClure cited analysis suggesting Anthropic added at least $5 billion in revenue in a quarter, but estimated a $150 billion to $200 billion run rate by the end of next year rather than the previously discussed $200 billion to $300 billion.

Ben Black argued that standard SaaS metrics such as ARR may fit poorly because Anthropic sells compute across models with different values and costs. He interpreted reported 900% net revenue retention as the same customer cohort spending nine times as much in the comparable three-month period a year later, but speakers questioned whether such growth and operating profits can persist as model prices fall and models become outdated. One speaker compared AI to cigarettes: free samples could attract users to a product that generates continuing demand. But the S-1’s reported AI spending of about 0.1% of total operating expenditure had an unclear scope, and panelists disagreed on what the figure meant. Low spending could reflect a shortage of useful applications, not simply high prices.

The adoption debate centered on whether AI demand is supply-constrained when providers are preordering compute while companies are still running pilots. Companies often have yet to connect AI to the data and systems needed for meaningful business results, and must also overcome security, implementation and organizational barriers. One panelist described the shift required: moving from employees acting as a “compiler” between AI inputs and outputs to orchestrating AI toward outcomes. McClure expected productivity benefits to reach ordinary companies as users move from basic prompts to agents, but the pace depends on organizational capacity to use expanding compute supply.

The panel also weighed competition and investment risk beyond Anthropic’s listing. OpenAI reportedly had 50 million subscribers among 900 million weekly users, while Anthropic’s consumer subscriber count was undisclosed. One view was that Apple could use its distribution and investment capacity once AI is good, safe and user-friendly enough; another favored on-device AI for most consumer needs while expecting complex enterprise workflows to remain cloud-based. On design software, speakers said Claude Design had raised concerns about Canva and Figma, but argued Figma’s collaboration network and existing role in companies’ workflows still matter. Wallach said investors should judge whether Anthropic can remain defensible and profitable over 10 to 20 years, while recognizing that competitors offering intelligence more cheaply or free could challenge its business.