Benchmark Invests in Manus: VC Math vs Geopolitics Math

About a month ago, I got early access to Manus, the AI agent product made by the Chinese startup, Butterfly Effect, that briefly went viral globally. A month later, Bloomberg scooped that Benchmark, a leading Silicon Valley VC firm, invested to fuel the company's growth. (I wrote an earlier post on how Manus stacks up against GPT and Gemini as a hypothetical junior analyst, if you are interested in a product review.)

The investment, if true, is of course notable. It is a major fundraising event for one of the few AI agent products that caught consumer traction. It also comes at a time when the US-China decoupling intensifies in all dimensions. For the most part, VC investing between the two countries has already shut down completely – a trend I called out at the end of 2023. At that time, all major VC firms that had operations in both countries had to split operations and branding completely, a so-called “splinter VC” phenomenon, due to US government scrutiny. Yet, Benchmark, which does not have any non-US operations (notwithstanding an ill-fated expansion effort to London back in 2000), flies in the face of all macro trends.

Benchmark's bet on Manus is both contrarian and rational, illustrating how “VC math” and “Geopolitics math” can work both against and for each other.

VC Math

VC math, by definition, is forward looking and consistently adheres to the “power law” of investing. Simply put, a top-tier early stage VC looks for startups that can generate 100-times return to pull the entire fund’s performance up. Meanwhile, startups in the same fund that fail and go to zero is also a perfectly ok outcome. To achieve this outcome, the VC has to identify the right startup, be able to invest in the startup, and (this element mostly gets overlooked) own a big enough stake in this startup, so when it does become a 100x juggernaut, the VC’s ownership translates to a large return and not get diluted away.

This math only works for disciplined early-stage VCs, not ones who raised larger funds for late stage investing or cross-over hedge funds. Benchmark is one of the few disciplined VCs, who has stuck to their knitting as a top-tier early stage investor and never raised a mega-fund for later stage investments, despite all the temptations to do so. It also stuck to its high bar of a large ownership stage, usually at the Series A stage, to ensure that its model works in VC math.

This Manus deal looks to be a trademark Benchmark investment. If the rumored numbers are roughly true – $75 million investment for a $500 million valuation – I’d venture to guess that Benchmark’s investment is about $60 million. The rest of the $15 million would come from follow-on investments from Butterfly Effect’s seed round funders – ZhenFund, HongShan (fka Sequoia China, a victim of splinter VC), and Tencent. A $60 million check into a post-money $500 million valuation gives Benchmark a 12% stake. If the valuation is pre-money, then the ownership stake is 10.4%. Guess how much did Benchmark own in Uber after its Series A investment there? 11%! 

This deal is right in the range of what Benchmark would consider an investment with good “VC math”. This $500 million valuation is, in fact, quite cheap compared to what other AI startups in Silicon Valley are getting. Cognition AI, who makes the coding agent, Devin, is valued at $4 billion. Anysphere, who also makes an AI coding platform, Cursor, is rumored to be raising at a $10 billion valuation. To be sure, Manus (or Monica, Butterfly Effect’s first AI chatbot product) does not have nearly the revenue traction as Devin or Cursor. But it is also aiming at a much larger consumer space, not software engineering as a business vertical use case, so there is more room to grow (or dream). Benchmark, with past home run successes in the consumer space (e.g., Snap, Pinterest), may see Manus’s potential and be better positioned to materialize it than most investors.. And when you have pre-product, pre-revenue, and talent-only startups like Mira Murati’s Thinking Machines Lab fetching a $10 billion valuation, Benchmark’s investment in Manus looks like a steal. 

(This is not an assessment of whether Thinking Machines Lab will succeed or not. Murati and the team she assembled is the crème de la crème of AI, so it is hard to bet against the company.  But from a VC math perspective, it’ll require a $30 billion outcome somehow for the investment to make sense, regardless of whether the company succeeds or not.)

My hunch is Butterfly Effect took a “geopolitical discount” due to the US-China decoupling and tension, a case where “geopolitics math” is working in favor of “VC math”, if there is a VC firm willing to take the deal and the additional risks associated with it.

Benchmark took the deal…and shouldered the risks.    

Geopolitics Math

Does VC math have to adhere to the rules of geopolitics math? Yes and no.

Geopolitics math, by definition, is backward-looking, remedial, and changes all the time. Take the US export control math as an example. Ever since the first round of export control in October 2022, there have been multiple subsequent rounds that redefine, expand, and broaden the scope and computation plus bandwidth power when it comes to semiconductor products and equipment that the US government deems unacceptable to be sold to China. That math is still fluid as we speak; further changes to the AI diffusion rules are in the works and anticipated.

US investments into China is another example of changing geopolitics math. When I was serving in the Obama administration, CFIUS was actively scrutinizing acquisitions by Chinese companies of US wind farms or other real properties that might be geographically close to US military bases. Fast forward more than ten years later to now, just about every US investment into any Chinese tech startups are ex-ante scrutinized for their dual-use implications into military, surveillance or other national security related use cases, so much so that no VC firm would bother any more. This recent trend is what makes the Benchmark Manus deal an anomaly on the surface. Even though Manus is just another consumer AI agent trying to help users do research or book flights, it takes very little effort in today’s political climate to draw some loose connections from the Butterfly Effect to the Chinese government or military. Geopolitics math is fickle and shifts when the geopolitical wind changes direction. 

So is that a problem for Benchmark? Of course it is. If Manus gets that level of scrutiny from DC, the deal is toast and the investment goes to zero! 

But is that a big enough problem in terms of VC math? Not really. In early stage investing, startups go to zero all the time. Even though the reason why Manus may go to zero is a bit unsatisfying if geopolitics is the cause, a zero is a zero. Life goes on. 

On the other hand, not investing in Manus could mean missing out on the next ByteDance of the AI age. Startups easily take 10 years or more to mature and scale to IPO level. There is no way to predict which way the geopolitical wind will blow even next month, let alone 10 years from now. If you are true to your craft as an early stage VC, it is a shot on goal worth taking.

Benchmark’s consistency and focus on its early stage investing craft is many decades in the making. Reading Bill Gurley, now a former GP of Benchmark, explains the firm’s investment in Webvan, an online grocery delivery website that blew up in the first dotcom bubble, you can see the same internal logic spinning among the partners today that led to its investment in Manus. A firm like Benchmark “does not succeed by avoiding failures.” 

The one risk that may make Benchmark regret this deal is reputational. Touching any Chinese AI startup, seen as a politically toxic asset, as an American investment firm will invite criticism, however baseless. When this deal was reported by Bloomberg, some of Benchmark's competitors quickly accused the firm of helping “the enemy.” Many of these detractors are sadly VC in name only and beltway lobbying firms in reality. 

Will Benchmark's reputation be tarnished, unfairly or not, by this deal, or will its enduring brand as one of the best early-stage VCs prevail? Benchmark has survived a lot of high-profile VC-startup dramas over the years. Webvan was a high-profile failure. Its marriage with Uber had a lot of drama (and literally a TV drama that came out of it). And then there was WeWork, another high-profiled, dramatic Benchmark portfolio company with its only TV adaption. Through it all, the firm maintained incredible consistency, discipline, and never raised a big fund. It is currently investing out of its 11th fund, a modest $425 million dollar fund raised in mid-2024 focused on AI, to do what it has always been doing – early stage Series A venture investing. 

If you combine the realities of VC math with geopolitics math, Benchmark’s investment in Manus is the economically rational thing to do. The world could use some more economic rationality these days.