IDC's latest study of the Open Compute Project's market footprint, released Tuesday morning at the OCP Global Summit in San Jose, places worldwide spending on OCP-recognized IT infrastructure at approximately $800 billion by 2029 — a figure 2.7 times larger than the $295 billion the same study projected for that year just 12 months ago.

By 2030, IDC projects spending rises further to $892 billion, with OCP-recognized equipment accounting for an estimated 51 percent of all spending across the infrastructure categories the study covers. The revision, which IDC attributes primarily to AI investment, compresses what was once a decade-long buildout story into a five-year surge running at a compound annual growth rate of 42 percent, according to the 2026 IDC Market Impact Study.

The study was commissioned by the OCP Foundation, the non-profit trade association that publishes open hardware specifications for servers, racks, power, and cooling — a standard disclosure applicable to any market research commissioned by an industry body. IDC is one of the three major IT research firms globally, alongside Gartner and Forrester.

The announcement came at the OCP Global Summit 2026, the foundation's annual technical conference, which opened Monday at the San Jose McEnery Convention Center and drew an estimated 7,000 to 9,000 hyperscaler engineers, ODM hardware teams, and data center architects for four days of sessions focused on AI-ready, high-density infrastructure. The summit continues through Wednesday. Details are available on the 2026 OCP Global Summit event page.

"AI is changing what the industry needs from its infrastructure, and these findings show the growing role of open designs in meeting that demand," said Cliff Grossner, Ph.D., Chief Innovation Officer at the OCP Foundation and author of the study announcement. "Our Community is bringing companies together to solve challenges that span the entire data center, from silicon and servers to cooling, power and the grid. That collaboration helps organizations put new technology to work at scale."

What a 2.7× Revision in 12 Months Actually Means

The December 2025 edition of the same IDC study projected OCP-recognized infrastructure spending would grow from $132 billion in 2025 to $295 billion in 2029, at a compound annual growth rate of 22.2 percent. The OCP December 2025 study set those figures as the near-term baseline.

The 2026 edition, commissioned again by the OCP Foundation, renders those figures obsolete before they could be acted on. The new 2025 baseline is set at approximately $154 billion — already above the prior study's near-term projection—and the trajectory from $154 billion to $892 billion over five years implies a rate of acceleration that the industry's own methodologies were not calibrated to anticipate.

IDC projects that roughly $628 billion of OCP-recognized infrastructure spending in 2030 will be AI-related, representing approximately 70 percent of the total forecast. That concentration means the data center infrastructure conversation and the AI hardware conversation have effectively merged: building a data center now means building for AI, and OCP specifications have become the dominant reference standard for how that hardware is designed.

GPU Servers Are Driving $814 Billion in Server Spending by 2030

Servers represent the dominant spending category in the forecast, with GPU-accelerated systems driving the projection. By 2030, IDC projects server spending alone within the OCP-recognized segment will reach about $814 billion—more than the entire OCP market total for 2029. That implies switching, rack, power, and cooling hardware collectively represent only a fraction of the 2030 figure, with server compute so dominant that secondary infrastructure spending is almost a rounding error in the headline number.

The technical reason for GPU server dominance is architectural: the matrix multiplication required to train and run large language models parallelizes efficiently across GPU cores in a way that general-purpose CPUs cannot match. A single GPU-accelerated server node today may draw 5 kilowatts or more from the rack, and fully provisioned AI training clusters routinely push individual rack power loads past 100 kilowatts—a density conventional air-cooled data center design cannot accommodate.

IDC's October 2025 Worldwide Quarterly AI Infrastructure Tracker, a separate study not commissioned by OCP, projected that globally — across all vendors and specifications, not just OCP-recognized equipment — accelerated servers would comprise more than 95 percent of total AI infrastructure spending by 2029, also at a compound annual growth rate of 42 percent. The United States was projected to grow AI infrastructure spending at 40.5 percent annually, China at 41.5 percent, underscoring the geographic breadth of the buildout.

Beyond servers, the new OCP study forecasts meaningful growth in switching equipment, rack infrastructure, power, and cooling as organizations build the dense, high-throughput systems required for increasingly demanding AI workloads. Liquid cooling, in particular, has become a central design challenge rather than an optional premium: direct-to-chip cooling systems — in which chilled water circulates through cold plates mounted directly on GPU and CPU packages — can support rack power densities exceeding 150 kilowatts, compared with a ceiling of approximately 35 to 50 kilowatts for well-designed air-cooled facilities. Industry figures suggest liquid cooling now accounts for approximately 34 percent of new AI data center builds, up from roughly 8 percent three years ago.

Open Specs Spread Further Than OCP Membership

The OCP Foundation's influence extends well beyond its formal membership base — a dynamic central to understanding why IDC's forecast is so large. An enterprise or cloud operator that never joins OCP can still purchase OCP-recognized servers from a Taiwanese ODM (original design manufacturer) built to OCP specifications. The specification functions as a de facto industry standard that propagates through the supply chain independently of membership status. IDC explicitly identifies the growing ecosystem of solution providers and the broad availability of openly shared design specifications as key factors accelerating adoption.

That membership base is itself growing rapidly: 165 new members joined OCP during 2026, following 127 in 2025 and 83 in 2024 — an acceleration that tracks closely with the AI infrastructure surge pulling suppliers, operators, and startups into open collaboration. The OCP Marketplace now lists 709 products and solutions, and active engineering work spans 224 project groups across 23 formal projects.

The membership growth pattern matters not just as a metric but as a signal about the supply chain. When 165 companies joined OCP in a single year, they joined to help write the specifications that will define AI data center hardware through the rest of the decade — not simply to purchase equipment someone else designed. That level of industry engagement allows OCP's open specification model to stay ahead of a rapidly shifting technical requirements landscape.

What Is OCP-Recognized Infrastructure — and Why Does the Forecast Now Exceed IDC's Own Prior Total AI Market Projection?

The $800 billion figure for 2029 deserves a precise definition. "OCP-recognized infrastructure" covers OCP-compliant equipment across servers, external storage, network switching, IT racks, power systems, cooling infrastructure, carrier-grade networking, and related peripherals — a broader category than the colloquial "AI servers" conversation suggests.

Within that broader definition, the OCP-recognized market is now projected to represent 51 percent of covered global infrastructure spending by 2030 — meaning OCP-based designs, originally pioneered by hyperscalers including Meta, Microsoft, and Google to strip cost and inefficiency from proprietary hardware, are on course to define the majority of all covered infrastructure spending globally, not just within the hyperscaler segment.

The comparison to IDC's own prior AI infrastructure market sizing is arresting: IDC's October 2025 Worldwide Quarterly AI Infrastructure Tracker — using a narrower definition focused on AI compute and storage — projected total global AI infrastructure spending across all vendors would reach $758 billion by 2029. The new OCP-specific forecast of $800 billion for that same year, for OCP-recognized equipment alone within a broader category, now exceeds IDC's own prior projection for the entire global AI infrastructure market.

The definitional difference explains part of the gap — the OCP study counts a wider set of infrastructure categories than the AI Infrastructure Tracker —, but the comparison makes concrete what the 2.7× revision otherwise implies in the abstract: the open hardware ecosystem, within a single year of AI demand acceleration, has grown beyond what the industry's own analysts thought the total addressable market would be.

What Engineers Are Building at the OCP Summit This Week

This week's technical agenda reflects the specific engineering problems behind the forecast numbers. Sessions center on GPU platform design, advanced thermal and liquid cooling, 48-volt and high-voltage DC power delivery, and modular data center architectures—the four engineering domains where GPU cluster buildouts are hitting physical constraints that require new specifications rather than incremental refinements of existing ones.

At April's OCP EMEA Summit in Barcelona, the foundation announced the Open Rack Wide (ORW) specification, a major expansion of its rack ecosystem for AI workloads. The ORW addresses a specific physical limitation: the standard Open Rack v3 form factor was designed before GPU trays reached their current width and depth requirements, and the specification needed to expand to accommodate next-generation GPU system designs without sacrificing the interoperability guarantees that make open specs valuable.

At the same Barcelona event, OCP formalized a Current/OS DC microgrid alliance to develop open standards for direct-current microgrids within data centers — extending the foundation's specifications into high-voltage DC power distribution, which eliminates AC-to-DC conversion losses at each server's power supply and meaningfully improves overall data center energy efficiency at scale.

"With no end in sight to the very large AI data center buildout, the collaborations and innovation standardizations within the OCP Community are more important than ever," George Tchaparian, CEO at OCP, said at the Barcelona event.

For hardware OEMs and ODMs, the revised IDC forecast quantifies the scale of demand for open-spec AI infrastructure through the end of the decade. For hyperscalers and neoclouds accelerating their GPU cluster buildouts, it signals that the open design ecosystem will supply most of the infrastructure they need — and that the engineering decisions being made at this week's Summit in San Jose will define what that infrastructure looks like.

Frequently Asked Questions

What is the Open Compute Project, and why do its market numbers matter to companies that haven't joined it?

The Open Compute Project Foundation is a US-based nonprofit trade association that publishes open hardware specifications — design blueprints for servers, racks, power systems, and cooling — under licenses that allow any manufacturer to build compliant equipment without paying a royalty. Founded in 2011 by Meta, Microsoft, and Google to strip proprietary overhead from hyperscaler hardware, OCP has grown to 602 corporate members. Its market influence extends far beyond that membership because ODMs in Taiwan and elsewhere produce OCP-specification equipment sold to operators who never formally joined.

When IDC counts "OCP-recognized infrastructure," it measures how much global data center hardware is built to these open specifications—a figure that now represents the majority of covered infrastructure spending by 2030. Any company buying, designing, or investing in AI data center hardware is operating in a market where OCP specifications are effectively the dominant reference standard.

Why did the IDC forecast for 2029 more than double in a single year?

The December 2025 IDC study projected $295 billion in OCP-recognized infrastructure spending by 2029, using a compound annual growth rate of 22.2 percent from a 2025 baseline of approximately $132 billion. The 2026 study raises the 2025 baseline to $154 billion — already above the prior study's projection — and extends the CAGR to 42 percent through 2030, driven by AI investment that was either not yet committed or not yet reflected in order books when the prior study was conducted.

AI-related spending is projected to reach $628 billion of the $892 billion 2030 total, meaning roughly 70 percent of OCP infrastructure growth through the end of the decade is attributable to AI workloads specifically. The combination of a higher starting point and a nearly doubled growth rate produces the 2.7× revision.

What does liquid cooling have to do with the OCP forecast surge?

Liquid cooling is not a peripheral detail—it is the physical constraint the entire AI server buildout is organized around. GPU-accelerated servers for AI training now routinely push rack power densities past 100 kilowatts per rack. Air cooling reaches a practical limit of approximately 35 to 50 kilowatts per rack in well-designed facilities. Above that threshold, air simply cannot move heat fast enough without unacceptable energy waste and infrastructure complexity.

Direct-to-chip liquid cooling, which routes chilled water through cold plates directly mounted on GPU and CPU packages, supports densities exceeding 150 kilowatts per rack. NVIDIA's current-generation GB200 NVLink rack design specifies liquid cooling as a requirement, not an option. When IDC projects $814 billion in OCP-recognized server spending by 2030, those servers predominantly require liquid cooling infrastructure to operate — which is why the cooling, rack, and power categories in the OCP study are growing alongside the server line, not independently of it.

Does the $800 billion OCP forecast include all AI infrastructure spending globally?

No — and the comparison to IDC's own prior figures clarifies why this matters. IDC's October 2025 Worldwide Quarterly AI Infrastructure Tracker, using a narrower definition focused on AI compute and storage hardware across all vendors, projected total global AI infrastructure spending would reach $758 billion by 2029.

The new OCP Market Impact Study's $800 billion figure for 2029 covers a broader set of infrastructure categories (including rack, power, and cooling alongside servers and switching), but applies only to OCP-recognized equipment within those categories. The fact that an OCP-specific subset now exceeds IDC's prior estimate for the total global AI infrastructure market signals the scale of the revision underway: the open-hardware ecosystem has grown to encompass more spending than the entire AI infrastructure market was projected to represent just one year ago.