|
Over 300 organizations rely on The Information for exclusive insights into public and private companies, including in-depth analysis of the ways in which tech moves markets. Click here to contact our corporate and enterprise team to learn more. Hi! Natasha here, with Kalley, Steph and Rocket. It says something about the hype around ex-OpenAI executive Mira Murati’s new startup, Thinking Machines Lab, that when Sequoia Capital’s Alfred Lin said his firm was “talking” to the startup, that was deemed worthy of a headline. We’ll do you one slightly better: Murati has told people she’s also had conversations with Vinod Khosla, an early OpenAI investor, about possibly investing, according to one of those people, though it’s unclear if the firm will actually invest. But Murati might have good reason to put off clinching a funding deal: attracting new recruits. Thinking Machines has a strategy to make joining the firm extremely lucrative. The company, which was just incorporated in December, has been telling candidates that it will grant stock options with strike prices near zero and that they can exercise those options nearly immediately after joining, according to a person who has spoken to the company’s employees. This means recruits would have to pay very little to exercise their options if they joined the company and won’t have to worry about the options expiring. That may explain why the company in the last month told some recruits that it does not yet have a 409a valuation, the internal share price estimated by an outside consultant, according to the person who spoke with the company’s employees. Spokespeople for Khosla Ventures and Thinking Machines Lab declined to comment. This arrangement could help the startup avoid a dilemma. Investors have been discussing investing $1 billion in the company at a roughly $10 billion valuation including the money, we have heard. Such a price, unusual for a young company, helps founders retain control. But it’s not necessarily a selling point for job candidates, who prefer to get shares in a startup when prices are cheap—and have plenty of room to grow. Without an established valuation, the startup seems to be able to offer early candidates equity at low prices—while fielding investment offers at much higher prices, which will make those candidates instantly wealthier on paper after the round is done. Exercising options early—thereby buying shares at a very low price—could also help employees save on taxes if they sold those shares in the future, since options are taxed based on the difference between the strike price and the value of shares when options are exercised, compensation consultants and lawyers tell us. It’s not unprecedented for new startups to grant options before a company has set a 409a valuation, these advisers say. Since the strike price of stock options is generally determined by the fair market value of a company’s shares, companies sometimes hold off on receiving written offers from investors, even if they have discussed or verbally committed to a certain valuation, until their negotiations have finished. What’s unusual in this case is the price difference. A seed-stage startup that offers candidates options at a strike price near zero, and then raises a round that values it at $20 million, is offering candidates some upside. But one that offers ultracheap equity and then raises at a $10 billion valuation is in a different ballpark altogether. The situation speaks to the white-hot nature of AI investments. Safe Superintelligence, a less than year-old startup started by former OpenAI chief scientist Ilya Sutskever that also has no launched product or revenue, is in talks to raise at a $30 billion valuation, Bloomberg reported. At the same time, all these startups are in a cutthroat race for talent, which explains why OpenAI has brought in Thrive Capital, a major investor, to make the pitch to employees that leaving the ChatGPT maker would be a bad financial decision. Murati’s strategy seems to be working. Thinking Machines Lab has attracted several prominent OpenAI alumni, including cofounder John Schulman and researchers Barret Zoph and Luke Metz. At the time of its launch, the startup had 29 employees, more than two-thirds of which hail from Murati’s former employer, OpenAI. While many are likely excited about working with an elite group of researchers and alongside a founder with many ties to top Silicon Valley VCs and large tech firms like Microsoft, the prospects for huge financial gains don’t hurt. And they’ll mint some of that wealth once its new valuation hits. Now over to Natasha for news on another former CTO… I have the scoop on a new startup founded by former Plaid Chief Technology Officer Jean-Denis Greze and former Google artificial intelligence and machine learning director Tony Vincent. San Francisco-based Town, which connects small businesses with tax advisory services, has raised $18 million, in a financing led by First Round, with participation from AI-focused venture firm Conviction Capital, Jack Altman’s Alt Capital and WndrCo, as well as angel investors such as Mercury CEO Immad Akhund and Quora CEO Adam D’Angelo. Launching a tax services startup seems particularly ambitious given the recent, abrupt shutdown of Bench, a 600-person startup that sold accounting services to over 10,000 customers, mostly small businesses and startups. After Bench’s bank called in its venture debt, it told customers they’d have to find a new provider. Then a new buyer scooped it up in a fire sale. That episode exposed a problem in the business model of fintech startups that offer an alternative to services, like bookkeeping and taxes, which had previously been the domain of local businesses. While self-service software was part of their pitch, they also employed hundreds of professionals to provide custom services—keeping costs high. Greze says he’s trying to avoid the pitfalls that hung up startups such as Bench that were launched before the rise of generative AI. Right now, the months-old startup has 10 employees and will look to grow to 13 this year—and plans to stay that way until it hits “the right level of automation.” Greze envisions that Town will pair human advisers with clients for professional advice, but expects to automate 90% of the administrative work that goes into preparing a tax return, for instance. “We feel really strongly that for something as important as your taxes, you want a human to look at it, but that can make the human 90% to 95% more efficient,” he said. Town is also using artificial intelligence internally to do customer support, such as auto-drafting responses to clients before a human sends an email, and to look through customer’s tax returns to find ways to save money. Town uses a number of foundational models, such as OpenAI and Anthropic, as well as some of the top open-source large language models, such as Meta’s Llama. Greze is an example of how founders are again trying to transform high-touch services businesses like accounting or tax preparation. This time they’re using generative AI, which they say will help them write their own code faster and automate human tasks, keeping costs low.
| 
