IDC: 2029 OCP Infrastructure Spending Forecast Raised to $800 Billion, Nearly Tripling Last Year's Estimate
nashnova research
IDC revised its 2029 forecast for global OCP-certified equipment spending from $295 billion to roughly $800 billion — a near-threefold jump that signals AI infrastructure investment is outrunning even the industry's most optimistic projections from a year ago.
Why did the same number nearly triple in one year?
IDC's 2029 OCP-certified equipment forecast stood at $295 billion last year. The new figure: roughly $800 billion — an upward revision of about 2.7×.
This means → the shift is not about gradual demand beating expectations. The *acceleration itself* exceeded what forecasting models assumed.
In plain terms = the speed that looked "fast" a year ago now looks slow — the forecasters couldn't keep up with the pace they were forecasting.
Where does the money go — and who captures the biggest share?
Servers are the single largest category, projected to account for roughly $814 billion in 2030. GPU-accelerated systems — servers purpose-built for AI workloads — are the primary growth driver.
AI-related spending is expected to reach roughly $628 billion of the 2030 OCP infrastructure total, or about 70%.
This means → growth in OCP-certified equipment is nearly synonymous with growth in AI compute. The remaining 30% is largely pulled along by AI-driven buildouts.
Switching gear, rack infrastructure, power, and cooling also grow with data-center expansion, but at far smaller scale than servers.
OCP's ambition goes beyond equipment — what does "chip to grid" mean?
OCP (the Open Compute Project — an industry alliance pushing open, non-proprietary designs for data-center hardware) is extending its scope in both directions: upstream to chips, downstream to the power grid.
Over the past year, OCP members advanced work on high-power rack design, liquid cooling, Ethernet-based AI networking, and open chiplet specifications.
In plain terms = OCP used to govern "what goes inside the rack." Now it wants to govern "where the power comes from, where the heat goes, and how the chips snap together" — open standards across the full chain.
A majority share by 2030 — is that number credible?
IDC projects OCP-certified equipment will account for roughly 51% of total spending in the infrastructure categories it tracks by 2030 — crossing the majority threshold for the first time.
The compound annual growth rate from 2025 to 2030 is projected at roughly 42%, starting from approximately $154 billion in 2025.
This reflects a shift in open design from "an option for hyperscalers" to "an industry default." But whether 42% compound growth can hold through a likely supply-constrained period around 2027 is the key test of this forecast.
What is OCP planning next?
Support multiple AI chip types coexisting on open platforms, rather than locking into a single GPU architecture.
Drive cross-layer design coordination within data centers — designing chips, servers, cooling, and power as an integrated system, not separate silos.
Extend deployment scenarios from hyperscale data centers to enterprise, regional, and edge environments — what OCP calls the "AI compute continuum."
This means → OCP is betting that AI compute demand will eventually spill beyond large data centers. Whoever lands open components across the widest range of settings captures the standard-setting power.
市场有风险，内容仅供研究参考，不构成投资建议。