Bloomberg Talks · · 16 min
Partner at Lightspeed Venture Partners Sebastian Duesterhoeft Talks Investing in Anthropic
With Sebastian Duesterhoeft
ChorusBring the market's conversation to your AI. Learn more
Summary
Lightspeed Venture Partners partner Sebastian Duesterhoeft discusses the firm's early investment in Anthropic, made when the company had under 300 employees, roughly two years of history and under $100 million in revenue. He credits the decision to Anthropic's enterprise focus, a strong research team and alignment on trust and safety, and says the deal was contrarian at the time. He highlights Anthropic's compute strategy, which he says was shaped by capital constraints and only became widely understood through the S-1 filing, with a fungible compute pool serving training, inference and internal needs. He argues that Anthropic's safety emphasis is commercially aligned because enterprises need trustworthy models. On the wider market, he expects multiple large language models, including open-source models such as Mistral (a Lightspeed investment) alongside closed models, to coexist, particularly for sovereign and data-specific needs. He sees AI transforming every industry over the long term while accepting that valuations will become overheated in places and that the path will not be linear.
Key takeaways
- Sebastian Duesterhoeft (bullish on Anthropic): Anthropic's enterprise focus, strong research and mission alignment made it a compelling early bet
- Sebastian Duesterhoeft (bullish on Anthropic): Anthropic's safety and commercial goals are aligned long term, since enterprises need trustworthy AI
- Sebastian Duesterhoeft on Open-source AI models: Open-source and closed-source AI models will coexist; it is an 'and', not an 'or'
- Sebastian Duesterhoeft (bullish on Mistral AI): Mistral's local European model still matters for sovereign and critical infrastructure needs
- Sebastian Duesterhoeft (bullish on Anthropic): Anthropic's compute strategy, shaped by capital constraints, is one of its core strengths
What was said, by company and theme
7 views from 1 speaker. Every quote is checked against the episode's transcript (how).
Anthropic
Across the shows: 144 speakers in other episodes, 253 bullish and 117 bearish calls, past 30 days
Where everyone stands on Anthropic
Sebastian Duesterhoeft · Lightspeed Venture PartnersBullish
Anthropic's enterprise focus, strong research and mission alignment made it a compelling early bet
Lightspeed invested in Anthropic when it had under 300 people, was about two years old and had under $100 million in revenue. The firm was drawn to its enterprise focus, since Lightspeed is an enterprise investor first, and to a research team it saw as very strong. It also saw strong alignment with the company's mission and values on trust and safety. The investment was not consensus at the time, and the speaker argues the best investments tend to be controversial.
“the focus on enterprise. We are enterprise investors first and foremost... saw a team that was incredibly strong on the research side, but also one that we felt we had a lot of alignment with when it comes to the mission and the values of the company.”at 3:57
Sebastian Duesterhoeft · Lightspeed Venture PartnersBullish
Anthropic's safety and commercial goals are aligned long term, since enterprises need trustworthy AI
The speaker argues that dual-use technologies have always been this way, citing fertilizer and diamonds, so the safety focus is important for a powerful technology. He says that in the long term the commercial and safety/risk profiles are very aligned, because enterprises that use the technology need it to be trustworthy. He therefore sees Anthropic's safety emphasis as commercially supportive, not a drag on profits.
“I think it's really important. This is an incredibly important and powerful technology... And I also think in the long term, both commercial and sort of safety and risk scores are actually very aligned.”at 6:18
Sebastian Duesterhoeft · Lightspeed Venture PartnersBullish
Anthropic's compute strategy, shaped by capital constraints, is one of its core strengths
The speaker says compute strategy is one of Anthropic's core strengths, one that was not well understood until the S-1 filing brought it to light. He attributes it to constraints: for long periods Anthropic had less capital than others in the industry, which made it more creative in how it managed compute, and that approach is still helping it today. He also describes Anthropic's compute as a fungible pool used for training, inference, internal use and other purposes, with real scale effects as the model reaches a larger customer base.
“I really think it is one of the core strengths and maybe one that maybe even as recently as the S1 wasn't really understood... I think it was constraints... This was a company that for long periods of time had less capital than others in this industry.”at 8:30
Mistral AI
Across the shows: 2 speakers in other episodes, 1 bullish and 0 bearish calls, past 30 days
Sebastian Duesterhoeft · Lightspeed Venture PartnersBullish
Mistral's local European model still matters for sovereign and critical infrastructure needs
Although the European 'LLM for Europe' idea is talked about less now, the speaker argues open source still has an important role. Sovereign angles matter, and critical infrastructure in Europe needs a local model. He also notes that customers who want to bring very specific data to models in ways they cannot with the closest closed model create a role for open models. He expects the European and US models to coexist. The host noted Lightspeed is an investor in Mistral.
“I think there's still really an important role for open source to play. Right? There is gonna be sovereign angles... it's still very important that there is a a local model that can be used by some of the critical infrastructure that exists in in Europe.”at 7:40
Open-source AI models
Where everyone stands on Open-source AI models
Sebastian Duesterhoeft · Lightspeed Venture Partners
Open-source and closed-source AI models will coexist; it is an 'and', not an 'or'
The speaker expects several large language models to thrive, as multiple hyperscalers have in the past. He calls competition healthy because it brings out the best in each company and is good for the world. He expects open-source models alongside closed-source models, saying this is not an either-or question but more of an 'and'. He describes customers' specific data needs and sovereign requirements as reasons for both to coexist.
“I think we've seen it in the past with hyperscalers where there are multiple... I think it's healthy to have competition... And we think there are gonna be open source models on closed source models. This is not an, you know, an or. It's it's more of an and.”at 7:12
AI: overall outlook
Where everyone stands on AI: overall outlook
Sebastian Duesterhoeft · Lightspeed Venture Partners
AI will eventually change every industry, role and job, though the path will not be straight
The speaker steps back and asks whether AI will change every industry, functional role and job over a long time. His answer is a clear yes. He acknowledges the path from today to that point will not be a straight line, but he sees the long-term direction as clear and believes the long-term view will be right.
“is it going to change every industry, every functional role, every job that exists out there? And when I ask myself that question... my resounding answer to that is yes. Does it mean that from where we sit today to that point in the future it's going be a straight line? No.”at 11:43
AI bubble & tech valuations
Where everyone stands on AI bubble & tech valuations
Sebastian Duesterhoeft · Lightspeed Venture Partners
Some AI companies will be overvalued along the way, but the long-term AI thesis will hold
The speaker expects pockets of overvaluation and some companies getting the difficult balance between spending and returns wrong along the way. He admits he will not predict those instances. He still argues that over the long term the AI thesis will be right, so temporary overvaluation is not a reason to reject the technology's long-run value.
“Doesn't mean we're going have pockets along the way that are overvalued... We're going to have those instances. I'm not going to predict them right... But over the long term I think we'll be right.”at 12:03
Across the shows
What other episodes Chorus covered said about the same companies in the 30 days up to this one.
- Anthropic144 speakers · 253 bullish, 117 bearish calls
- Mistral AI2 speakers · 1 bullish, 0 bearish calls
Topics
- AI safety and commercial alignment
- AI valuations and overvaluation
- AI's long-term economic impact
- compute strategy
- enterprise AI focus
- open vs closed model coexistence
- sovereign AI models
More from Bloomberg Talks
- Instant Reaction: Anthropic Rockets in Stock Market Debut, Breaks IPO Record · November 19, 2026
- Nasdaq President Nelson Griggs Talks Anthropic IPO · November 19, 2026
- Elizabeth Economy Talks APEC Summit · November 19, 2026
- Citi CEO Jane Fraser Talks Outages, Banking Industry · November 18, 2026
- NYC Comptroller Mark Levine Talks Anthropic IPO · November 18, 2026
Ask your AI about this episode
Chorus gives Claude, ChatGPT and other assistants every view from this episode and every other show Chorus follows. Ask who agrees, who changed their mind, and what else was said about the same companies.
Chorus reports what speakers said. It isn't investment advice. How Chorus works · About Chorus