The AI Chip Rivalry: Why AMD's Bold Move Could Shake NVIDIA's Throne

The race to dominate artificial intelligence (AI) hardware has taken a dramatic turn. HSBC's recent analysis suggests AMD is no longer playing catch-up to NVIDIA in the high-stakes world of AI chips. Instead, its latest MI350 series GPUs are now positioned to challenge NVIDIA's Blackwell GPUs head-on—a shift that could redefine market dynamics and unlock a 40% upside for AMD's stock.

Technological Parity: AMD's Performance Leap

AMD's MI350 series, particularly the MI355X model, has closed the gap with NVIDIA's Blackwell B200 in critical performance metrics.

HSBC's report highlights two key advantages:
1. Memory Bandwidth: The MI355X delivers 22.1 terabytes per second (TB/s), nearly triple the Blackwell's 8 TB/s. This is vital for training large language models, where data throughput often bottlenecks performance.
2. Cost Efficiency: AMD's chips command a 30% price discount at $25,000 versus NVIDIA's Blackwell. While NVIDIA's sparse TFLOPS (5,000) still edge out AMD's dense TFLOPS (3,000–4,000), the narrowing gap combined with pricing power creates a compelling value proposition.

This technical parity is no small feat. For years, NVIDIA's CUDA ecosystem and proprietary architectures have been near-impenetrable. AMD's progress here signals a potential inflection point.

Market Dynamics: A New Frontier in AI Hardware

The stakes are high. NVIDIA currently commands over 90% of the discrete GPU market, with hyperscalers like AWS and Azure deeply entrenched in its ecosystem. AMD's challenge isn't just about hardware—it's about cracking a fortress.

HSBC's analysis underscores three critical shifts:
1. Revenue Surge: AMD's AI revenue is now projected to hit $15.1 billion by 2026, a 57% beat over consensus. Even a 10% market share gain could cost NVIDIA $5 billion annually.
2. Pricing Power: AMD's ASP increase to $25,000 adds $3 billion to its 2026 revenue, a testament to its growing leverage.
3. Software Challenges: While AMD's ROCm stack is improving, NVIDIA's 5 million CUDA developers and rack-scale solutions (e.g., 72-GPU clusters) remain a hurdle.

AMD's stock has already risen 14.6% year-to-date, but HSBC argues it hasn't yet priced in this AI upside.

Valuation: A Bullish Case for AMD's Stock

HSBC's upgraded price target of $200—a doubling from $100—implies a 38.8% upside from current levels ($144). This reflects a 31x 2026 earnings multiple, aggressive but defensible if AMD's AI pipeline delivers.

The catalysts are clear:
- Q2 2025 Earnings: Gross margins and data center revenue will signal whether the MI350 is gaining traction.
- MI400 Launch (2026): AMD's next-gen chip aims to rival NVIDIA's Vera Rubin, though execution risks remain (e.g., reliance on Broadcom's Ethernet switches).

Critics like Goldman Sachs caution about valuation risks, but HSBC's analysis suggests AMD's AI trajectory justifies the optimism.

Risks and Roadblocks

The path isn't without potholes:
- NVIDIA's Ecosystem: CUDA's dominance isn't easily displaced. NVIDIA could lower Blackwell prices or accelerate software updates to counter AMD's cost advantage.
- Software Lag: AMD's ROCm stack needs to catch up in developer tooling and adoption.
- Supply Chain: The MI400's reliance on Broadcom's Ethernet switches introduces a single point of failure.

Investment Thesis: Buy AMD for a 40% Upside

AMD's move into AI isn't just a niche play—it's a full-scale assault on NVIDIA's crown jewels. The MI350's performance and pricing advantages, coupled with a $200 price target, make this a compelling “Buy” for investors willing to tolerate execution risks.

Actionable Takeaway:
- Target: $200 (40% upside from $144).
- Hold Until: 2026, with 2025 earnings and MI400 milestones as key checkpoints.
- Risk Management: Monitor NVIDIA's countermeasures and AMD's software progress.

In the AI arms race, AMD has fired a shot heard around the industry. The question now isn't whether it can compete—but whether it can finally conquer.