Meta to Spend Billions on AMD Gear, AI Scare Trade Continues

2026-02-24 · Source: Bloomberg Tech · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Software Development & Engineering · Depth: Intermediate, extended

Summary

Meta has agreed to a multi-billion dollar deal with AMD to acquire chips and data center supplies, including AMD's MI450 accelerators, over the balance of the decade, potentially totaling 6 gigawatts of capacity. This move, which includes share warrants for Meta, aims to diversify Meta's AI infrastructure beyond Nvidia and its own custom chips, focusing on AI inference workloads. Concurrently, the market is experiencing "AI whiplash" as Anthropic's new agentic AI tools, like Claude, are causing significant shifts in software company valuations. For example, IBM's shares dropped after Anthropic suggested Claude could modernize COBOL, while other partnerships, such as with Intuit, have boosted stock. Startups like MatX are also raising substantial capital, over $500 million, to develop specialized LLM hardware to compete with Nvidia, emphasizing computational density and hybrid HBM/SRAM designs. Basis, an AI accounting tools startup, secured $100 million at a $1 billion valuation, highlighting the need for domain-specific AI solutions.

Key takeaway

For CTOs and VPs of Engineering navigating the rapidly evolving AI landscape, your strategy should prioritize diversifying AI hardware procurement to reduce reliance on single vendors like Nvidia, as demonstrated by Meta's AMD deal. Simultaneously, evaluate the potential for agentic AI tools to disrupt existing software ecosystems, focusing on domain-specific applications that offer guaranteed accuracy and tailored user experiences to avoid market "whiplash" and secure competitive advantages.

Key insights

Diversification in AI chip procurement and specialized AI agents are reshaping market dynamics and software valuations.

Principles

• AI infrastructure demand is insatiable.
• Domain-specific AI tools offer distinct advantages.
• Hybrid memory architectures can optimize LLM performance.

Method

AI chip startups are developing custom hardware by combining HBM and SRAM to achieve high throughput and low latency for large language models, breaking compatibility with previous chip generations for optimized LLM workloads.

In practice

• Diversify AI hardware suppliers to mitigate vendor lock-in.
• Invest in domain-specific AI solutions for critical workflows.
• Explore hybrid memory architectures for LLM deployment.

Topics

• AI Chips & Infrastructure
• Agentic AI
• Large Language Models
• AI in Accounting
• Data Center Energy Costs

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Investor, AI Product Manager, AI Engineer

Related on AIssential

• Deal between AMD and Meta could bring greater innovation to GenAI market — Strategic "chips-for-stock" partnerships are diversifying AI compute supply and challenging single-vendor dominance.
• Abishai Financial Asia Eyes AI Memory Supercycle - Analytics Insight — Agentic AI architectures are driving unprecedented demand for high-bandwidth memory, creating a "supercycle" in semiconductor equities.
• What to Expect from Nvidia This Week — Nvidia is expanding its AI hardware and ecosystem, while AI agents and models face both market adoption and regulatory challenges.
• 615: Nvidia Q4, Ben Evans on AI Economics, Meta + Google TPUs, China Solar at 1TW/Year, Anthropic vs DoD, AWS Outages, METR, Sora, and Marty Supreme — Diversifying AI compute suppliers and architectures is crucial for large-scale AI development, despite integration challenges.
• 📈 Data to start your week — Global market data reveals shifts in solar energy, US economic growth, developer resource usage, and AI hardware/valuation dynamics.
• AMD Soars on Blockbuster AI-Fueled Forecast — AI demand is driving significant growth and strategic shifts across semiconductor, cloud, and consumer tech sectors.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Bloomberg Tech.