Top image (representative): Data center server racks. Photo: Carl Lender / CC BY 2.0 via Wikimedia Commons

If you want to know where investors think AI’s value will settle, follow the money into the plumbing. One recent week of venture rounds tracked by AlleyWatch (30 notable U.S. deals totaling $2.5 billion in the week ending Oct. 3) included four large South Bay and Peninsula rounds. Our analysis of third-quarter venture totals covers the quarter-wide trend toward fewer, larger checks. This piece looks at the individual deals closest to home. None was for a consumer app. All four were for the layers underneath AI: interconnects, compute, edge chips, and security.

The rounds

• GMI Cloud, Mountain View, $223 million. Cloud infrastructure for AI workloads, including GPU computing, model inference, and agent runtime services. Founded in 2023 by Alex Yeh, it has raised $238 million in total, with backers including Nvidia, according to AlleyWatch.
• Armadin, Palo Alto, $255.5 million. An AI-native cybersecurity platform that finds and validates exploitable risks. Founded in 2025 by David Slater, Evan Pena, Kevin Mandia, and Travis Lanham, it has raised $444.7 million in total from investors including Accel, Andreessen Horowitz, Kleiner Perkins, and Menlo Ventures.
• SiMa.ai, San Jose, $150 million. Hardware and software for deploying “physical AI” at the edge. Founded in 2018, it has raised $505 million in total.
• CScale, Palo Alto, $145 million. Optical interconnect technology for AI infrastructure. Founded in 2023 by Sanjai Kohli, it has raised $173.3 million in total, with backers including Intel Capital, Nvidia, Sutter Hill Ventures, and Valor Equity Partners.

The same list included a smaller Peninsula round: Redwood City’s Presto, which sells AI and automation tools to restaurants, raised $10 million, bringing its total to $192.6 million.

Reading the pattern

Three of the four large rounds address a constraint Nvidia’s Jensen Huang named publicly on Oct. 15. In an interview reported by Yahoo Finance, Huang said power, unfinished data centers, and components are limiting how fast supply can grow. Optical interconnects (CScale) address how fast data moves between chips. GPU clouds (GMI Cloud) rent out scarce compute. Edge silicon (SiMa.ai) moves some inference out of the data center entirely. Nvidia appears on the investor lists of two of them, so the chipmaker is also helping fund the ecosystem that buys its chips.

The fourth, Armadin, reflects the other side of the AI build-out: as more operations become automated, the attack surface grows too. Its total of nearly $445 million for a company founded last year shows how much capital is chasing AI-era security.

The long view

Every computing era rewards a few application companies and many more in infrastructure. The mainframe had its peripheral makers and the internet had its routers and optical networks. The AI era’s equivalents are being funded right now, a short drive from where the original Valley grew up. The open question is timing. These rounds closed as public-market investors were questioning debt-financed AI spending, and private valuations usually follow public sentiment with a lag. The South Bay is betting on the long cycle.

Note: Funding figures are as reported by AlleyWatch’s weekly funding report and have not been independently confirmed with the companies.

Sources

• AlleyWatch, “The Weekly Notable Startup Funding Report: 10/5/26”
• Yahoo Finance, “AI Stocks Plunge, but Jensen Huang Says Nvidia Gets New Chip Orders Almost Daily,” Oct. 15, 2026