The money trail runs through Phoenix, and it ends at a chip company that says it can beat Nvidia at its own game. Jack Selby, a partner at Thiel Capital, has been investing in AI chip startups, including the custom inference chipmaker Etched, through a network of investors and limited-partnership structures built around his ties to Arizona, according to people familiar with the investments.
The structure is as interesting as the target. Selby, who has become one of the most active figures in Arizona’s technology scene, has been pooling capital from wealthy individuals and family offices in the state and routing it into early-stage technology companies through vehicles designed to give limited partners a direct line into deals that normally go to Silicon Valley funds. The approach is a regional answer to a national problem: how to get into the best AI deals when the capital is concentrated on the coasts.
Etched is the flagship of the strategy. The company builds custom chips specialized for running AI inference, the process of generating answers from trained models, and its architecture is designed around the transformer models that dominate modern AI. The company’s pitch is straightforward: a chip built for one job runs that job far faster and cheaper than a general-purpose processor, and for inference workloads, the specialized chip wins on both measures.
The ambition goes further. Etched’s founders have said their design directly challenges Nvidia’s CUDA software ecosystem, the platform that has made Nvidia’s chips the default choice for AI developers by making them easy to program. A custom chip without CUDA faces an uphill fight against a decade of developer habit, but Etched argues that inference does not need the flexibility CUDA provides, and that the efficiency gains justify the migration.
The Nvidia question is the center of the industry’s attention. Nvidia’s dominance rests on the combination of hardware and software, and its market value has made it the most valuable chip company in history. Startups like Etched represent a bet that the AI workload is becoming specialized enough to support dedicated silicon, the same transition that produced Google’s tensor processors and Amazon’s Trainium chips.
Selby’s Arizona network gives the investments a particular character. Arizona has emerged as a technology hub in its own right, with chip plants under construction, a growing startup scene and a political climate that has attracted investors from California. The state’s limited partners have money to deploy and a preference for doing business with people they know, a combination that Selby has used to assemble funds quickly and without the overhead of a traditional venture firm.
Thiel Capital’s involvement ties the deals to one of the most influential investors in technology. Peter Thiel, the PayPal co-founder and Palantir chairman, has backed a range of contrarian technology bets over two decades, and his firm’s interest in AI chips is consistent with a thesis that the industry’s value will migrate to the physical layer of computing.
The Arizona angle has a history behind it. Thiel himself relocated to the state and has been active in its political and business communities, and Selby’s network has grown out of those connections, drawing in investors who might never have been reached by a California fund. The result is a pipeline that connects Southwestern capital to Silicon Valley technology, a flow that has become more visible as Arizona’s chip industry has expanded.
The structure of the Arizona vehicles is unusual enough to draw scrutiny. People familiar with the arrangements said the limited-partnership structures give investors exposure to individual deals rather than a diversified fund, concentrating risk in exchange for a direct share of the outcomes. That shape suits investors who know the companies well, but it is a different proposition from the diversified approach most family offices take.
The company’s technical claims will be tested by its customers. Inference chips live or die on benchmarks and total cost of ownership, and Etched’s early hardware has been evaluated by a small circle of cloud providers and AI developers, people familiar with the industry said. A win with a major customer would validate both the company and the investors who backed it; another delay would test the patience that the Arizona structure was built to provide.
Etched’s fortunes will test the thesis. The company has raised significant capital and shipped early hardware, but it faces the same challenges as every Nvidia challenger: manufacturing capacity, software tooling and the patience of customers who have built their systems around the incumbent. The Arizona money gives it runway; the market will decide whether the runway leads anywhere.
The broader signal is the direction of capital. Money is flowing into custom AI silicon from every direction, from the largest cloud providers to the newest startups, and the Arizona deals are part of that current. Whether the challengers win or lose, the industry’s dependence on a single chip supplier is being tested as never before.
For Selby’s network, the Etched bet is a way to participate in the AI story without paying the valuations of the public market. The structure keeps the investors close to the companies, the deals out of the headlines and the potential returns, if the thesis holds, ahead of anything a diversified portfolio could offer. The strategy is regional in origin and national in ambition, and the chips it backs are aiming at the center of the industry.
