Will private AI company valuations be reset soon?We perhaps boldly asked on Saturday if the Trump Pump, AI-hype combo has led to overvalued public tech companies like Nvidia, shouldn’t it be the same-same for private VC-backed AI companies? Due to more recent events and further processing, we wanted to share a couple of deeper thoughts on the subject. As a follow-on to the $500 billion Stargate announcement, SoftBank is in talks to invest up to $25 billion in OpenAI for a whopping rumored valuation of $300 billion! If this is true, this would be a huge 73% uptick from the $157 billion valuation Open AI fetched in October 2024, when the company raised $6.6 billion. SoftBank also bought $1.5 billion in OpenAI shares from existing investors a few weeks back, presumably at a more favorable valuation. This potential investment would make SoftBank OpenAI's largest single backer, surpassing Microsoft's current investment of around $13 billion.
Why Masayoshi Son would want to almost double OpenAI's value the week after Nvidia got a half-a-trillion market cap haircut is not believable, and our bet is he is back to the negotiating table with some adjusted deal math today. Masa, who was born in Japan but is a third-generation Zainichi Korean, is widely regarded as one of the most influential and successful tech investors globally, with an estimated net worth of $34 billion.
Growing into your market valueIn our last post, we suggested that the Nvidia/DeepSeek drama was a wake-up call on several levels, and it might be a good time to review the private company AI valuations while they are still private. In particular, we suggested starting with a sales projection versus value analysis, and that it might be smart, after seeing the success of DeepSeek’s open source AI model, to start thinking of AI as more of a commodity services model with potentially smaller profit margins than anticipated. We are currently finishing the final Cryptonite 300 top private Web3 companies list, and so far, we have identified 75 AI unicorns amongst the batch, 36 of which were newly minted in 2024. Of those, the top 30 are listed below.
Market Value-to-SalesLet’s use OpenAI’s current $157 billion valuation as a case study and speculate on a $300 billion new funding valuation in relation to sales and sales growth. OpenAI posted $3.7 million in revenue for 2024, and typical figures say OpenAI is projecting $11.6 billion in revenue in 2025. As one data point, it took Microsoft, Google, and Facebook each three years and Apple four years to go from $3 million to $11 million in sales. If OpenAI hits its 2025 mark, the company will be the fastest-growing tech company in history. A related data point is the Price/Sales ratios (i.e., total company valuation divided by projected/forward sales). Only Microsoft (14.4x ) has a better P/S ratio than OpenAI’s current valuation (13.5x, assuming record sales), with Google (4.4x), Meta (7.3x), and Apple (8.8x). At Masa’s proposed $300 billion valuation, OpenAI’s P/S ratio goes to 25.9x.🤔 Another interesting analogy here is the case of ByteDance, the owner of TikTok. As of 2024, ByteDance's valuation was estimated at $300 billion against $146 billion in sales in 2024. ByteDance’s 2025 forecast has not been published, and their future is very mirky given the forced sale of TikTok, but the company has been operating with a 2.5x P/E.
Downward pressure on salesThere are several reasons that OpenAI and the private venture-backed companies may not meet their revenue goals, including radical competition from BigTech players (Microsoft, Facebook, Google, Apple) and 75 private company AI unicorns and counting (xAI, Anthropic, DeepSeek). Google is bracing for what industry analysts are forecasting as a significant drop in the company’s US search market share from 92% to below 50% by 2025 😳 primarily due to the rise of generative AI search competitors. After a 10 percent cut in ‘managerial positions’ in December 2024, Google has introduced a voluntary buyout program for employees in its ‘Platforms and Devices’—Android, Chrome, etc., to streamline operations and focus on artificial intelligence development. Google’s only way to pull out of its potential revenue spiral is to compete fiercely in the AI app markets.
A recent study by Menlo Ventures also showed that OpenAI lost market share in enterprise AI this year, dropping from 50% to 30%, while private company competitor Anthropic doubled its share in the sector. Behind Anthropic are dozens of startups nipping at OpenAI’s (and each other’s) heels with faster, cheaper, and more specialized AI services. As an example, at Cryptonite, we prefer Perplexity’s general search and xAI’s Grok for a more technical search, including most of the financial numbers and analysis for this post. We are waiting for xAI’s laptop version, and then we can share our xAI links as well. It’s a humble reminder that Google Search was the 17th search engine brought to market; Facebook successfully followed the Friendster flame-out and MySpace. So, any talk about the benefits of OpenAI’s ‘first-mover advantage’ runs contrary to Silicon Valley entrepreneurial company history.
Open source is here to stay and will be a commoditizing forceWe still strongly contend that, in the long run, AI’s biggest winners will not be the companies that produce remarkably fast marginal cost commodity services but the AI-powered users leveraging the new tools to spend less time producing a lot more work. The biggest disruption of the DeepSeek story is not that the company is going to win in the market—it won’t because, like the current TikTok, it must adhere to the CCP’s censorship guidelines. It’s because DeepSeek’s novel, ‘optimization advances,’ as Zuck calls them, has shown Meta and Sam Altman what is possible and what code to swipe from the Chinese company to do the same.
All this code-swiping has inevitable long-term effects for all AI companies— cheaper development costs, the need for fewer AI chips, and commodity pricing for general-purpose consumer and enterprise GenAI services. DeepSeek's open source approach fosters a collaborative and sharing community that drives code enhancements, makes fixes and offers new solutions—all at accelerated rates and a fraction of the cost of closed models. The open-source strategy also offers several cost-efficiency advantages over its closed source rivals, such as eliminating hefty licensing costs and offering lower API fees. Meta chief AI scientist Yann LeCun has said the biggest takeaway from DeepSeek's success is the value of open source AI models versus proprietary ones. LeCun says, 'DeepSeek's achievements are not that China is surpassing the US in AI, but rather that open source models are surpassing proprietary ones.' Our view is that the open source model will dominate the AI sector like never before, and most of the first GenAI companies will need to adjust as Zuck and Sam are doing.
Just as we thought the ultimate AI company winners were coming into view, new open source AI models are rewriting the sector all over again. Open-source AI companies and their communities will play an increasingly significant role as AI innovators, accelerators, and cost cutters, spurring open access to advanced AI capabilities. All this new activity is also challenging current propped-up VC portfolio company valuations. Has the new reality that AI VCs may have invested too much money at valuations that are too high set in yet? Time shall tell—and tell us shortly. —Anthony Perkins, founder & editor Cryptonite The Trump Doctrine on full displayWe highly recommend watching this Megyn Kelly interview with the new US Secretary of State, Marco Rubio, where Mr. Rubio elaborated on the ‘America First’ doctrine while also acknowledging the shift from a unipolar to a multipolar world order. He also touched upon China, Greenland, and Panama.
After the interview, Rubio met with Panamanian President José Raúl Mulino regarding the Chinese Communist Party's growing control over the Panama Canal's operations. He expressed to President Mulino that China’s influence poses a threat to US rights and violates the 1977 Panama Canal Treaty, and the US might take ‘necessary measures’ if immediate changes are not made. President Mulino characterized the conversation as ‘respectful’ and ‘constructive’ and shortly after announced that Panama would not renew its participation in China's Belt and Road Initiative upon its expiration.
So far, the Trump Doctrine is 3-for-5. Colombian President Gustavo Petro initially refused to accept US military flights carrying deported Colombian migrants; however, he quickly backed down to pressure and offered his presidential plane to fly to the US to bring them home. Meanwhile, Richard Grenell, President Trump's envoy for special missions, visited Venezuela and met with President Nicolás Maduro. Mr. Grenell came home with six freed American hostages and President Maduro’s (whose ‘election’ the US still does not recognize) agreement to take back all deported Venezuelan nationals, including members of the notorious Tren de Aragua gang. The third victory was the Panama deal, as outlined above. Meanwhile, the tariffs on goods from Canada and Mexico became effective this weekend. It appears that soon departing Canadian Prime Minister Pierre Trudeau and new Mexican President Claudia Sheinbaumdo have no interest in addressing the Boogeyman’s requests for stricter enforcement against Mexican cartel drug smuggling, fentanyl production and distribution, and undocumented immigration across the Mexican and Canadian borders into the US. Given the reasonable nature of the US requests, we see both Presidents Trudeau and Sheinbaumdo backing down soon. The world will quickly figure out that President Trump’s negotiating bluster to ‘take back the Panama Canal, buy Greenland and make Canada the 51st state’ is just his way of getting everyone’s attention. Whether you love or hate the Boogeyman, his negotiating tactics work, and that is very good news for the US—and Canada, Mexico, Greenland, and Panama, too. |