Silicon Valley is the most expensive place in North America to buy a kilowatt.

Why Silicon Valley still commands a premium

Silicon Valley is where a large share of the internet's peering, cloud on-ramps and enterprise headquarters sit within a few miles of each other. For workloads that serve West Coast users, sit adjacent to a cloud region, or depend on dense private interconnection to partners already in Santa Clara, no other western market replicates the ecosystem.

What has changed is growth. CBRE reported 489.2 MW of inventory at the end of 2025, up only 4% year over year, while primary markets as a group grew 36%. The constraint is power and buildable land, not demand.

Source: CBRE on Silicon Valley and CBRE North America Data Center Trends H2 2025. Figures describe the metro market, not a specific facility.

Pricing, and why the range is so wide

CBRE puts Silicon Valley asking rates at $180 to $275 per kW per month, against a national average of $195.94 for a 250 to 500 kW requirement. That $95 spread is the widest published band of any primary market, and it is not noise. It is the difference between a legacy cabinet row in an older Santa Clara building and contiguous space in a modern hall.

CBRE also notes that requirements of 10 MW or more of contiguous space command a premium here specifically because so little of it exists. Scale does not buy a discount in this market. It buys a shortlist of one or two buildings.

• Confirm whether the quoted rate is for existing space or a future phase.
• Silicon Valley Power applied a 4% rate increase effective January 2026, so model the utility line separately from the facility rate.
• Ask what the same operator charges in Sacramento, Reno or Phoenix for the identical requirement, then price the interconnection difference honestly.
• Cross-connect volume drives real cost here more than in any other market. Count them before comparing.

When Silicon Valley is the right answer

It is the right answer when the workload genuinely uses the ecosystem: heavy private peering, latency to Bay Area users measured in single-digit milliseconds, adjacency to a specific cloud region, or a partner that will only cross-connect in Santa Clara. Those requirements justify the premium because moving elsewhere breaks the architecture.

It is the wrong answer when the driver is simply that the company is headquartered nearby. Compare against Sacramento and Reno for the same West Coast coverage at a materially different rate, and against Hillsboro when subsea and Pacific Northwest routing matters more than Bay Area peering.

Risks to resolve before shortlisting

• Power ceiling: municipal utilities in Santa Clara and the surrounding cities have finite headroom. Confirm the building's allocation, not the campus story.
• Age of the estate: a meaningful share of available capacity sits in older buildings with lower design density. Check what the hall actually supports per cabinet.
• Expansion: with 84% of construction preleased, growth room is scarce. Contract it rather than assume it.
• Seismic and regulatory: California adds design, permitting and compliance requirements that lengthen any buildout.
• Rate direction: utility increases are landing on top of rising facility rates. Model both.

A facility-ready Silicon Valley brief

1. Committed power, density per cabinet, and cooling method.
2. Every cross-connect and the partner on the other end of it.
3. Latency targets to named user populations and cloud regions.
4. Whether the requirement can split across two buildings, and at what cost.
5. Term, escalator, renewal rate and utility pass-through mechanics.
6. The comparison market you would accept if the premium fails the test.

What has been announced in Silicon Valley

Announcements are not filings. Nobody has to build what they announce, the figure is the one the developer chose to publish, and the date moves. They are here because megawatts are what a requirement is measured in, and because the size of what is coming changes who will take your call. Every figure below was read on the page it links to.

Announced capacity is collected by hand from trade reporting and company statements, and each figure links to the report it came from. It is kept apart from the permit record above and never added to it: one is a document a jurisdiction issued, the other is a plan a company published. Neither is availability.

Every Silicon Valley facility we track

144 data centers across Silicon Valley, from the facility database we hold as a Bridgepointe partner. It is the layer that says whether a building exists yet, which no permit record answers directly.

Operators with the most facilities here: Digital Realty (24), Equinix (17), Vantage Data Centers (11), CoreSite (10), Csquare (8), Prime Data Centers (7), STACK Infrastructure (6), Evocative (5), Prologis (5), NTT Global Data Centers (4).

Where they cluster: Santa Clara (73), San Jose (33), Sunnyvale (7), San Francisco (6), Fremont (4), Rancho Cordova (4).

The largest buildings coming to this market

29 facilities here are under construction or planned. These are the ten largest by floor area, and a size on file is not a promise of a delivery date.

From the Signal facility database, which we hold as a Bridgepointe partner rather than as a public record, so these rows carry no filing link and street addresses are held back. A building existing is not the same as space being available in it: every facility above may be full, pre-leased, or closed to your size. Finding out is a call, not a lookup. See how we verify capacity.