AI Silicon Buyers Are Rebalancing Between GPUs and Custom Chips as Market Heads Toward US$361 Billion

Analyst view: The market's next phase will be decided by how efficiently chips serve AI workloads at scale. GPUs remain the largest chip type, but hyperscalers are adding custom accelerators, while demand for inference is spreading beyond data centers to PCs, vehicles, robots and industrial systems.

DataM Intelligence values the global AI silicon market at US$77.59 billion in 2025 and forecasts it to reach US$361.14 billion by 2035, a 16.64% CAGR from 2026 to 2035. North America accounted for 40.21% of the market in 2025. The report identifies GPUs as the leading chip type and custom AI ASICs and inference silicon as among the fastest-growing areas. datamintelligence.com

The commercial shift is visible in recent company announcements: customers are committing to large-scale GPU deployments, while chip designers and cloud companies are co-developing custom processors. At the same time, memory and advanced packaging capacity are becoming key considerations in bringing AI silicon to market.

Request Executive Sample: https://www.datamintelligence.com/download-sample/ai-silicon-market?kailas

Recent Developments:
- June 7, 2026 - NVIDIA and SK hynix announced a multi-year technology partnership to develop memory for AI infrastructure and align supply with NVIDIA's roadmap. The companies also plan to use AI platforms to support semiconductor design and manufacturing. SK hynix Newsroom
- June 24, 2026 - OpenAI and Broadcom unveiled Jalapeño, a custom processor designed for large-language-model inference. The companies said the chip moved from design to manufacturing tape-out in nine months, with deployment planned at gigawatt scale.

- July 22, 2026 - AMD and Anthropic announced plans for up to 2 gigawatts of AMD Instinct MI450 GPUs in Helios rack-scale systems, with the first gigawatt expected to begin deployment in the first half of 2027.

- August 31, 2026 - NVIDIA and MediaTek expanded their collaboration across cloud infrastructure, local AI computing and automotive. MediaTek will adopt NVIDIA's NVLink Fusion platform for custom XPU development; NVIDIA also announced a US$3.5 billion investment in MediaTek convertible bonds.

- October 8, 2026 - GlobalFoundries and TSMC agreed to establish U.S.-based production of silicon interposers for TSMC's CoWoS advanced-packaging ecosystem under a multi-year US$2 billion agreement.

- October 21, 2026 - NVIDIA and manufacturing partners announced plans to expand supply capacity. The company cited a targeted 2.5-fold increase in substrate capacity over three years and an Amkor plan to grow advanced-packaging capacity by more than 300% by the end of 2028.

- October 27, 2026 - AMD and Meta highlighted their expanded infrastructure partnership, including AMD Helios systems running at Meta and collaboration across processors, GPUs, networking and software. AMD also announced a US$1 billion investment in Meta's equity offering.

Regional Outlook:

United States:

North America's 40.21% share reflects the concentration of chip designers, hyperscalers and AI infrastructure buyers in the region. Recent activity points to a broader strategy: secure GPU supply, develop workload-specific silicon and expand domestic packaging capacity. In October, the U.S. and Japan agreed on a semiconductor manufacturing project in New York, with an estimated cost of about US$10 billion and construction scheduled to begin by the end of 2026.

Japan:

Japan's opportunity spans semiconductor manufacturing, materials, packaging and edge AI applications in automotive, robotics and industrial automation. The government's AI and semiconductor framework sets out more than ¥10 trillion in public support through FY2030, intended to encourage more than ¥50 trillion in public-private investment over ten years. That policy focus could strengthen Japan's role in the supply chain as demand expands for AI systems and advanced packaging.

Purchase This Exclusive Report: https://www.datamintelligence.com/buy-now-page?report=ai-silicon-market?kailas

Additional Benefits:
✅ Get Interactive Dashboard Access (With No Additional Cost)
✅ Three months of analyst support to answer queries related to the report.

Market Segmentation:

The report segments the market by chip type, processing type, deployment environment, form factor, application, end-use industry, business model and technology node. Its public report page does not disclose percentage shares for individual segments; the quantitative figures below are therefore market size and regional share, while segment positions reflect the report's stated rankings.

- By chip type: GPUs lead market value, supported by broad programmability and established software ecosystems. Custom AI ASICs are gaining ground where high-volume workloads justify the cost of designing dedicated silicon. NPUs serve lower-power endpoint and edge applications.

- By processing type: Training remains a high-value use for large accelerator clusters. Inference is the expanding volume opportunity, spanning cloud services, enterprise applications and edge devices. The shift places greater weight on cost per token, power efficiency and memory performance.

- By deployment environment: Data centers and cloud platforms represent the largest value pool. Edge and endpoint silicon is extending AI processing into vehicles, PCs, robots and industrial equipment, widening the customer base beyond hyperscalers.

Company Profiles

- NVIDIA: Combines GPUs with CUDA software, networking and rack-scale systems. Its expanded work with MediaTek and SK hynix spans custom accelerators, local AI computing and next-generation memory.

- AMD: Competes through Instinct GPUs, EPYC CPUs, ROCm software and Helios rack-scale systems. Partnerships with Anthropic and Meta show its push to secure large deployments and co-design systems for customer workloads.

- Broadcom: Provides custom accelerator and networking technology for hyperscale AI infrastructure. Its work with OpenAI on Jalapeño demonstrates the growing role of custom silicon designed around inference workloads.

- MediaTek: Brings custom silicon, SoC design and connectivity capabilities to AI platforms from cloud to edge. Its expanded NVIDIA collaboration connects those capabilities to custom XPUs, AI PCs and automotive systems.

Request a Customized Report According to Your Business Strategy: https://www.datamintelligence.com/customize/ai-silicon-market?kailas

Contact Us:
Sai Kiran
Business Development Manager
DataM Intelligence 4market Research LLP
6th Floor, M2 Tech Hub, Lalitha Nagar, Habsiguda,
Secunderabad, Hyderabad, Telangana 500039
USA: +1 877-441-4866
Email: Sai.k@datamintelligence.com

About DataM Intelligence
DataM Intelligence is a global market research and business intelligence firm delivering actionable insights across healthcare, pharmaceuticals, chemicals, energy, technology, food, and industrial sectors. Through syndicated reports, custom research, consulting, and competitive intelligence services, the company helps organizations identify growth opportunities, navigate market challenges, and make informed strategic decisions in over 50+ countries worldwide.

This release was published on openPR.