Intel and Google Cloud have expanded their strategic partnership to advance AI infrastructure, with Google deploying Intel’s Gaudi 3 AI accelerators across its cloud platform and the two companies collaborating on custom silicon development. The move represents Intel’s most significant design win in the AI accelerator market as it seeks to challenge NVIDIA’s estimated 80% share of the data centre AI chip segment.
The expanded agreement, announced this week through Intel’s newsroom, includes Google Cloud offering Intel Gaudi 3 accelerators as a service to enterprise customers and joint development of future custom chips built on Intel’s advanced manufacturing processes. Google will also continue deploying Intel Xeon processors across its infrastructure.
The partnership arrives as enterprises seek alternatives to NVIDIA’s H100 and H200 GPUs, which have faced supply constraints and premium pricing throughout 2024. Intel positions Gaudi 3 as offering competitive performance for large language model training and inference at lower total cost of ownership, though independent benchmarks have shown mixed results depending on workload specifics.
“This collaboration demonstrates the market’s appetite for diversified AI compute options,” said Patrick Gelsinger, Intel CEO, in the announcement. Google’s deployment provides Intel with a crucial validation point as it attempts to gain traction in a market where AMD’s MI300 series and custom chips from Amazon and Microsoft have also emerged as NVIDIA alternatives.
The custom silicon component of the partnership leverages Intel Foundry Services, the company’s contract manufacturing division. Google has previously developed custom tensor processing units (TPUs) with other foundry partners, but this agreement signals potential Intel fabrication of future Google silicon designs using the company’s Intel 18A or Intel 3 process nodes.
For Google Cloud, the partnership addresses two strategic priorities: offering customers diverse AI infrastructure options and securing long-term chip supply through multiple vendor relationships. The company’s cloud infrastructure division has lagged behind Amazon Web Services and Microsoft Azure in market share, holding approximately 11% of the global cloud market according to Synergy Research Group’s third-quarter 2024 data.
Market implications
Intel gains immediate credibility in AI accelerators through association with a hyperscale cloud provider, potentially opening doors to enterprise customers evaluating alternatives to NVIDIA. The company’s data centre and AI division, which reported $3.3 billion in revenue for the third quarter of 2024, has struggled with declining margins as customers shift spending toward AI-specific chips rather than traditional CPUs.
Google Cloud strengthens its competitive positioning by offering differentiated AI infrastructure, particularly important as enterprises increasingly select cloud providers based on AI capabilities and GPU availability. The partnership also provides Google with leverage in negotiations with NVIDIA whilst maintaining access to that company’s market-leading chips.
NVIDIA faces limited immediate threat from the partnership, given its technological lead and ecosystem advantages, but the collaboration represents the type of customer diversification strategy that could gradually erode market share if Intel delivers on performance and cost claims. The announcement follows similar partnerships between AMD and major cloud providers, suggesting the AI chip market is entering a more competitive phase.
AMD, which has positioned its MI300X accelerators as the primary NVIDIA alternative, now faces another well-funded competitor with hyperscale backing. The company’s AI chip revenue reached $1.5 billion in the fourth quarter of 2024, demonstrating significant customer appetite for alternatives.
What to watch
The partnership’s success will depend on Intel’s execution delivering Gaudi 3 chips at volume and meeting performance specifications, an area where the company has faced challenges in recent product cycles. Google Cloud’s customer adoption rates for Gaudi-based instances will provide early indicators of enterprise willingness to move workloads off NVIDIA infrastructure.
The custom silicon collaboration timeline and specific process node selection will signal whether Intel Foundry Services can compete with TSMC for advanced logic production. Any announcements of additional hyperscale partnerships for Gaudi accelerators would substantially strengthen Intel’s market position and indicate broader industry diversification away from single-vendor dependency.
The expanded Intel-Google partnership reflects the AI infrastructure market’s maturation beyond single-vendor dominance, with meaningful implications for enterprise AI deployment costs and chip supply diversity over the next 18 to 24 months.
