Executive Summary
- Bank of America designated NVIDIA, Broadcom, and AMD as premier AI computing investments, establishing price objectives at $300, $450, and $280 respectively
- Mistral AI secured $830M in debt financing to acquire approximately 13,800 Nvidia GB300 GPUs for a Paris facility, representing roughly $575M in semiconductor revenue
- Broadcom welcomed OpenAI as a chip design collaborator and revealed more than $100B in AI procurement commitments through fiscal 2027
- AMD’s data center revenue trajectory shows expansion from $17B in 2025 to $77B by 2028, according to Aletheia Capital projections
- Worldwide cloud infrastructure expenditure reached $110.9B in Q4 2025, marking a 29% year-over-year increase
Bank of America has elevated NVIDIA, Broadcom, and Advanced Micro Devices to premier positions within the AI compute landscape, establishing specific price objectives while mapping potential challenges for each semiconductor leader.
These strategic recommendations arrive amid accelerating global AI infrastructure investments. Research firm Omdia reports worldwide cloud infrastructure expenditure climbed to $110.9 billion during Q4 2025, representing a 29% year-over-year surge. Projections indicate a 27% expansion rate for 2026.
Bank of America established a $300 price objective for NVIDIA, calculated at 28 times projected 2027 earnings. The financial institution highlighted NVIDIA’s dominant position in AI compute and networking technologies as the primary rationale.
Cantor Fitzgerald maintained its Overweight stance and $300 valuation for NVIDIA after analyzing the company’s GTC conference presentations.
French artificial intelligence firm Mistral AI contributed to positive momentum. The company obtained $830 million through debt financing to construct a new data center facility near Paris, equipped with 13,800 of NVIDIA’s GB300 GPUs. Industry analysts calculate the procurement could generate approximately $575 million in semiconductor sales for NVIDIA.
Space technology venture Starcloud secured $170 million at a $1.1 billion valuation. The enterprise previously deployed an NVIDIA H100 GPU into orbit and plans a subsequent satellite mission featuring a complete GPU cluster later this year.
On March 16, NVIDIA unveiled its Space-1 Vera Rubin computing module, engineered to handle data processing in orbit rather than transmitting raw information to Earth. NVIDIA has yet to announce shipment timelines.
Broadcom Anticipates $100 Billion AI Order Pipeline
Bank of America assigned a $450 price objective for Broadcom, calculated at 26 times projected 2027 earnings. The institution emphasized double-digit earnings expansion and robust free cash flow generation.
Broadcom recently welcomed OpenAI as a chip design collaborator through a multiyear agreement to jointly develop 10 gigawatts of custom AI accelerators. The company maintains similar partnerships with Google, Meta, Amazon, Microsoft, and Anthropic for custom silicon development.
Executive leadership disclosed more than $100 billion in AI chip procurement commitments secured for fiscal 2027. AI semiconductor revenue projections place next quarter at $10.7 billion.
Broadcom also secured a five-year, $970 million agreement with the Defense Information Systems Agency and initiated volume deliveries of its Tomahawk 6 switch chip.
AMD Pursues Data Center Expansion Strategy
Bank of America established a $280 price objective for AMD, emphasizing AI expansion and CPU market share gains.
Aletheia Capital maintained a Buy recommendation with a $330 price objective. The firm anticipates AMD’s server CPU revenue will expand at a 45% compound annual growth rate between 2025 and 2028.
Data center revenue forecasts show growth from $17 billion in 2025 to $77 billion by 2028.
AMD and Celestica introduced the Helios rack-scale AI platform. AMD also finalized a multi-year licensing agreement with Adeia Inc., settling all pending legal disputes between the organizations.
Lisa Su received appointment to President Trump’s Council of Advisors on Science and Technology.
AMD voiced concerns regarding its client and gaming divisions due to escalating memory costs.
