AI Bay: The Case for a 700 mph Underground Hyperloop Around San Francisco Bay
Silicon Valley did not stay in Silicon Valley. The gravitational pull of talent, capital, research labs, and frontier AI companies has shifted the center of gravity northward into San Francisco and the broader Bay Area. What was once a peninsula of chipmakers and startups is now the operational and cultural headquarters of the artificial-intelligence era. A more accurate name for the region is already circulating among founders and investors: AI Bay.
Yet the physical infrastructure of AI Bay remains stubbornly twentieth-century. Traffic on the Bay Bridge, 101, and 280 routinely turns a thirty-mile trip into a two-hour ordeal. Housing costs have priced out the engineers, researchers, and support workers the industry needs. Existing transit—BART, Caltrain, ferries, and buses—is useful but limited in speed, capacity, and coverage. Robotaxis will help with the last mile, but they cannot solve the core problem of moving large numbers of people quickly between the major nodes of the region.
The solution is not more surface congestion or another incremental rail extension. It is an underground hyperloop: a near-vacuum tube system running at up to 700 miles per hour, with roughly ten stops that circle the entire Bay. Passengers would board at stations in San Francisco, Oakland, Berkeley, San Jose, Palo Alto, and other key locations, travel at airline speeds, and then complete their journeys by robotaxi, transit, bike, or foot. Existing local transportation remains relevant precisely because the hyperloop handles the long, high-speed leg.
Why Underground and Why a Loop
An elevated or surface system would face enormous right-of-way, environmental, and political obstacles in one of the densest, most expensive, and most seismically sensitive regions in the United States. Going underground largely sidesteps surface disruption, preserves views, and can be routed under existing freeways or utility corridors where possible. A full loop—rather than a single point-to-point line—creates network effects: every station is connected to every other station without transfers, and the system serves both the Peninsula and the East Bay symmetrically.
A realistic loop distance, drawing on earlier Hyperloop One conceptual work for a Bay Area system, is on the order of 100–120 miles. Ten stations spaced around that loop would put most major employment and residential centers within a short robotaxi or transit ride of a stop. At 700 mph cruise speed (with acceleration and deceleration), travel times between opposite sides of the Bay would drop to minutes rather than hours.
Who Builds It
Two specialized companies fit the bill. The Boring Company has demonstrated and is refining tunnel-boring technology aimed at dramatically lower costs than traditional subway construction. Recent project figures and company targets put realistic near-term all-in costs for its smaller-diameter tunnels in the range of roughly $15–30 million per route-mile in favorable conditions, with longer-term ambitions well below that. For a hyperloop-scale tube (larger diameter and vacuum-ready), costs would be higher, but the company’s continuous-mining machines, smaller cross-sections relative to conventional rail tunnels, and “porpoising” launch techniques still offer a step-change versus traditional methods that often exceed $500 million to $1 billion+ per mile in urban U.S. settings.
For the hyperloop technology itself—pods, magnetic levitation or air-bearing systems, linear propulsion, vacuum maintenance, and control software—an Indian firm such as TuTr Hyperloop (incubated at IIT Madras and commercializing work from the Avishkar student team) is a strong candidate. Indian groups have built test tracks, developed pods, and won early commercial freight contracts. Their cost estimates for hyperloop infrastructure have been in the range of $10–20 million per kilometer in Indian conditions (roughly $16–32 million per mile), significantly below many Western high-speed rail figures. Partnerships between a U.S. tunneling specialist and an Indian hyperloop systems developer could combine low-cost tunneling know-how with proven pod and propulsion technology.
Cost Estimates
Any figure is speculative because no commercial passenger hyperloop yet exists and because California geology, seismic requirements, environmental review, labor rules, and station complexity will drive costs higher than optimistic early projections. Still, order-of-magnitude ranges can be grounded in published data.
- Hyperloop One’s 2016 internal estimates for a ~107-mile Bay Area loop (tunnel or mixed) ran $9–13 billion, or $84–121 million per mile.
- Broader industry ranges for hyperloop have varied from Musk’s original ~$11–17 million per mile (optimistic, mostly elevated) to $40–60 million per mile in various studies, up to $100+ million per mile in complex urban or tunneled settings. Realistic underground costs today are often cited in the $70–150+ million per mile range once stations, vacuum systems, propulsion, safety systems, and contingencies are included.
- The Boring Company’s demonstrated and targeted tunnel costs are far lower for vehicle tunnels ($15–30 million per route-mile range in recent projects and statements, with longer-term targets of $3–10 million), but a hyperloop tube requires a larger diameter, continuous vacuum integrity, emergency systems, and higher precision. Scaling those methods upward still suggests meaningful savings versus conventional bored transit tunnels.
- Indian estimates (TuTr/Avishkar-related) have put full hyperloop corridors at roughly $10–20 million per km ($16–32 million per mile) under lower-cost construction environments.
Illustrative total for a 110-mile underground loop with 10 stations:
- Optimistic (aggressive cost reduction via Boring Company methods + efficient systems integration): $8–15 billion.
- Mid-range (blending published hyperloop estimates with tunneling efficiencies): $15–25 billion.
- Conservative (California urban/seismic realities, full stations, contingencies, soft costs): $25–40+ billion.
Stations themselves are expensive—each major underground station could run hundreds of millions—so the ten-stop configuration is a significant cost driver. These numbers remain far below the trajectory of California High-Speed Rail (Phase 1 alone now projected around $126 billion) and many conventional urban rail extensions that have exceeded $1–2 billion per mile in the Bay Area.
Operating costs would be dominated by energy for vacuum pumps and propulsion, maintenance of seals and systems, and staffing. Ticket prices could be competitive with driving or flying once ridership scales, especially if the system captures a large share of intra-Bay commuting and business travel.
Solving Housing (and More)
Speed collapses distance. A reliable 10–15 minute trip from Oakland or the East Bay to downtown San Francisco, or from San Jose to the city, makes living farther from the core practical without sacrificing career access. That expands the effective housing supply and reduces pressure on the most constrained neighborhoods. Combined with robotaxis for the final leg, the system turns the entire Bay into a single labor market at human scale.
Additional benefits include reduced highway congestion and emissions, greater resilience (underground and grade-separated), and a visible infrastructure moonshot that matches the technological ambition of AI Bay itself. Local transit and robotaxis become feeders rather than primary long-haul options.
The Political and Funding Path
Technology and engineering are necessary but not sufficient. A project of this scale requires a dedicated political initiative: a regional authority or joint powers agreement spanning the relevant counties and cities, streamlining of environmental review under CEQA and NEPA where possible, and a clear funding package. Sources could include:
- Federal infrastructure and technology grants or loans
- State bonding or cap-and-trade allocations
- Regional sales or parcel taxes (building on recent Bay Area transit measures)
- Private capital and value-capture mechanisms around stations
- Public-private partnership with The Boring Company, the hyperloop systems provider, and real-estate partners
The political lift is real—California has a long record of slow, expensive infrastructure—but the alternative is continued paralysis. AI Bay’s economic output and tax base are large enough to support a transformative investment if leaders treat it as essential infrastructure rather than a discretionary amenity.
Conclusion
The region that invented the microprocessor, the modern internet, and now frontier AI should not be limited by 1950s road geometry and 1970s transit. An underground hyperloop circling the Bay at 700 mph with a handful of well-placed stations is ambitious, expensive, and technically demanding. It is also the scale of solution that matches the problem. Pair The Boring Company’s tunneling capability with an experienced hyperloop systems partner, secure the political will and funding, and AI Bay can move its people as fast as it moves its ideas. The last mile will still belong to robotaxis and local transit. The long mile—the one that currently wastes hours every day—can belong to the tube.
AI Bay: The Case for a 700 mph Underground Hyperloop Around San Francisco Bay https://t.co/sviOjO0E16 @garrytan @paulg @vkhosla @pmarca @bhorowitz @karpathy @sundarpichai @AravSrinivas @sama @DarioAmodei
— Paramendra Kumar Bhagat (@paramendra) November 13, 2026
