AI startup Anthropic, founded in 2021 and led by former OpenAI employees Dario and Daniela Amodei, recently raised $3.5 billion in a funding round, valuing the company at $61.5 billion. AIM spoke with Debarghya (Deedy) Das of Menlo Ventures, the principal overseeing Anthropic’s investment, among others, to understand what drew them to Anthropic.
Explaining the underlying belief, Das said, “We thought if OpenAI can be this big, Anthropic can definitely be huge.” Anthropic’s annual recurring revenue (ARR) grew from $1 billion to $1.4 billion in just six weeks, which was unprecedented at the time.
AI startups are raising massive rounds often without a tangible product in sight. Safe Superintelligence (SSI), founded by former OpenAI chief scientist Ilya Sutskever, is a key example and is reportedly valued at a whopping $30 billion.
Deedy recognised SSI as a prime example of high-risk, high-reward venture behaviour. “Most funding happens in boom and bust cycles, but the trend line is upwards,” he noted, adding that while AI dominates today, future venture cycles could favour something else, like robotics or biotech.
He also mentioned that VCs are eager to make substantial bets on companies founded by former employees of OpenAI, Anthropic or DeepMind, as they believe these individuals might discover something that current labs might not.
Besides Sutskever, many high-profile employees have left these research labs to establish their own startups, including the latest venture, Mira Murati’s Thinking Machine Labs.
“In a private market, especially in early-stage venture capital, you barely have revenue numbers, and you’re making a bet on the team, the market and their ability to execute,” he added.
In a recent interview, Khosla Ventures CEO Vinod Khosla said AI investments may incur losses, but a few high-return outliers will carry the wins.
Frontier AI in India
One question remains: from an investor’s perspective, how rational does this venture risk appear in a country like India, which is only beginning to invest in a homegrown LLM?
Das believes that India faces a considerable challenge in competing with Western labs. Although not entirely impossible, he admits, as an investor, he is unlikely to make such a bet.
“If you look at AI in India today, the market size doesn’t quite justify the level of R&D investment needed to make it a success,” he said.
Das views localised AI models, particularly those supporting Indian languages and voice applications, as a natural and achievable focus area for India’s AI ecosystem.
However, in an earlier post on X, he agreed that foundational models are not as effective in non-English languages, which is where India’s AI opportunity also exists.
Notably, if Western labs are so close to AGI and singularity, why do they continue hiring for posts despite emphatic warnings about automation?
In a recent interview, Anthropic CEO Dario Amodei said that he believes AI will handle 90% of coding in less than six months. This aligns with warnings from various AI labs regarding the near-total automation of coding, most recently highlighted by OpenAI CPO Kevin Weil.
Yet, the career pages of these AI labs feature numerous engineer postings.
To this, Das said, “When Anthropic says software engineers, large parts are going to be automated, they mean boilerplate stuff.” He added that for more complex systems, like a Google-scale search engine, AI isn’t close to replacing human expertise yet.
When asked if Menlo Ventures uses AI, Das responded, “I 100% use it to make my memos, and..to read the memo. If it can replace my investment decision-making, by all means, tell me what to pay for it.”
He pointed out that the essence of VC is collaborating with founders, which AI can’t replace.
Would Menlo Fund Another Anthropic?
“The next person who attempts to do exactly what Anthropic did is unlikely to succeed,” Das said, noting that this was reasonable if the approaches were distinct.
Startups exploring new methodologies, whether through new architectures or advances in interpretability, are instead well-positioned to attract funding.
Last year, Menlo Ventures and Anthropic launched the Anthology Fund, a $100 million initiative to support startups building applications using Anthropic’s technology. Other labs and startups like OpenAI and Perplexity also have startup funds exclusive to them.
At this point, the never-ending debate continues: are wrapper AI companies merely features, or can they evolve into sustainable businesses?
From a venture capital perspective, the distinction may matter less than it seems. As Das said, “Fundamentally, products are about—can you get users? Do they like you enough to stay? And will they pay you? And if those three things are true, I don’t care if it’s a wrapper or not.”
Das’ primary criticism of wrappers, however, is that they must provide substantial value beyond what’s available directly through ChatGPT, Claude, or other available LLMs. If not, they will struggle to attract and retain users.
He points to Cursor and Perplexity as success stories on the application layer.
The Era of Lean Startups
Some startups today are incredibly lean, with just a handful of people or even a single founder generating millions in revenue.
OpenAI’s Sam Altman discussed the future of billion-dollar startups in the intelligence era, which could be just one person and 10,000 GPUs. The concept of ‘one-person billion-dollar startup’ is not new and has been spoken about by Marc Andreessen and Ben Horowitz before.
AIM has covered this trend in the past. Cursor reached $100 million ARR with 20 employees, Midjourney hit $200 million with 11, Bolt.new made $30 million in four months with 20, and Lovable earned $10 million in 60 days with 15.
However, Das stressed that not all industries can have lean teams yet. “While AI-powered and cloud-based startups thrive with small teams, some industries still require large workforces, such as enterprise SaaS.”