Speaker · Updated
Lo Toney: latest market views
Plexo Capital
Summary
Lo Toney is the Founding Managing Partner of Plexo Capital, which he incubated and spun out from GV (Google Ventures). He is also a CNBC contributor and a former Comcast Ventures partner.
Chorus has 6 views from Lo Toney across 2 episodes on CNBC Closing Bell since October 9, 2026. Their latest calls: bullish on Anthropic and OpenAI.
Where they stand
Their latest bullish or bearish call on each company.
What they talk about most
The latest views
Newest first. Every quote is checked against the episode's transcript.
AnthropicBullish
Anthropic's public benefit fund is a positive step addressing AI disruption and public anxiety.
Toney, an early Anthropic investor, says the 5% equity donation to a public benefit fund is an important step on a disruption many expect and that people do not yet understand. He says models are becoming powerful enough to need controls, and the fund's philanthropic side and possible guardrails are both needed. He concedes the optics of regulatory capture, but says Anthropic's constitutional AI approach is why he invested and that there is no monopoly on proposals.
“Think it's an important step to begin to address what many expect will be a disruption”at 26:42
AI infrastructure & data centers
AI compute demand is still there; the risks are hyperscaler payback and rising cost of capital.
Toney says the AI selloff reflects CoreWeave's cost of capital, which rising interest rates are making more expensive. He says upcoming earnings must show that demand is still there, which he believes it is, and that hyperscalers can pay back their spending. If financing costs slow the buildout, compute could become scarce even with demand intact, which would hurt revenue growth.
“things that we wanna see are indications that the demand is still there. I believe it's still there.”at 32:02
OpenAIBullish
The OpenAI revenue gap looks like a gross-versus-net accounting issue, not lost customer demand.
Lo Toney says the roughly $20 billion difference reflects reconciliation between gross and net revenue, not vanishing demand. Operating signals are encouraging: enterprise sales reportedly more than doubled since July. OpenRouter data suggests OpenAI reached equal spending share with Anthropic in September, and its advertising business is making progress. Clearer reconciliation is needed, and AI revenue must also be judged on growth and what remains after distribution and inference costs.
“I don't see this disclosure as evidence that something is fundamentally wrong with OpenAI... that 20,000,000,000 difference doesn't mean 20,000,000,000 of customer demand just vanished and disappeared.”at 28:06
AI stocks are priced for perfection, so a revenue-number confusion can trigger a sharp sell-off.
OpenAI and Anthropic are central to expectations supporting the entire AI buildout, and some argue the global economy. Stocks are priced for exceptional execution and are priced to perfection. Confusion over a revenue number can trigger a sharp reaction, so if OpenAI even hints at a sneeze, the market might catch a cold.
“They've really become central to the expectations supporting the entire AI AI build out... stocks are pretty much priced for exceptional execution, they're priced to perfection.”at 29:27
AnthropicBullish
Anthropic's enterprise focus, team, and cost-efficient approach made it a compelling investment.
Plexo invested in Anthropic after getting direct access and a better look than at OpenAI. Anthropic's focus on enterprise, its team, and Dario's structured approach to models that use data more efficiently and need less compute were compelling. Cost was going to be a factor, and at an $18 billion valuation it was the firm's most expensive deal, which needed a trillion-dollar exit to make a venture return.
“we liked Anthropic's focus on the enterprise. We liked the team. We could tell that cost was gonna be a factor. And at that time, we thought Dario's plan of focusing on a more structured approach”at 30:39
Market should judge AI margins on cost per task, not cost per token, and frontier price cuts expand demand.
Open-weight and frontier models need similar infrastructure, so compute costs are the same. Margins depend on how token costs are structured. Open-weight models may need more tokens to complete a task, so cost per task completed is the better measure. Frontier labs cut prices when the next generation arrives, which lets the market expand to more users and volume, a Jevons effect. Margin transparency is needed, and the margin worry for Anthropic and OpenAI as they go public is warranted.
“all tokens are not created equal. And I think what we're seeing the market move towards is instead of cost per token, cost per task completed”at 33:24
Shows
- CNBC Closing Bell · 2 episodes
Ask your AI what Lo Toney thinks
Chorus gives Claude, ChatGPT and other assistants every view Lo Toney has shared on the shows it covers, next to everyone else's. Ask where they stand, what changed, and who disagrees.
Chorus reports what speakers said. It isn't investment advice.