Key Takeaways
- Advanced Micro Devices (AMD) is intensifying its challenge to NVIDIA with its new MI350 series AI accelerators, using competitive pricing to gain market share.
- The MI350 series, particularly the MI355X variant, demonstrates superior performance on key metrics, offering up to 40% more tokens per dollar in inference tasks compared to NVIDIA’s B200.
- Analyst forecasts for AMD’s 2025 AI chip revenue have been significantly increased, with some estimates reaching $15.1 billion, driven by adoption from major cloud providers like Microsoft.
- AMD’s market share in AI accelerators has grown from under 5% in 2023 to an estimated 12% in the first half of 2025, although NVIDIA still dominates the market with its CUDA software ecosystem.
Advanced Micro Devices (AMD) is intensifying its challenge to NVIDIA’s dominance in the artificial intelligence (AI) chip sector through targeted enhancements to its Instinct MI350 series accelerators, including a notable price recalibration that underscores growing market confidence in its competitive edge. This move, aligned with broader industry shifts towards diversified AI hardware supply, could propel AMD’s data centre revenues to new heights in 2025, amid escalating demand from hyperscalers and enterprise adopters.
AMD’s MI350 Series: Technical Specifications and Market Positioning
The MI350 series, unveiled in mid-2025, represents AMD’s latest foray into high-performance AI accelerators designed for data centre inference and training workloads. Built on a 3nm process node from Taiwan Semiconductor Manufacturing Company, the MI355X variant features 288GB of HBM3e memory and claims up to 1.8 times the inference performance of its predecessor, the MI325X. These specifications position the MI350 directly against NVIDIA’s Blackwell architecture, particularly the B200 GPU, which has faced production delays and high demand constraints since its announcement in 2024.
Recent adjustments to the MI350’s pricing structure reflect AMD’s assessment of its value proposition in a market where AI chip costs are under scrutiny. Industry reports indicate that the unit price has been set at approximately USD 25,000, a level that allows AMD to capture margins comparable to premium offerings while undercutting rivals on a performance-per-dollar basis. This pricing enables AMD to target cost-sensitive segments, such as cloud providers scaling AI infrastructure, where NVIDIA’s Blackwell units are priced around USD 35,000 per unit.
Comparative Performance Metrics
To contextualise the competitive landscape, consider key benchmarks from independent testing. As of July 2025, the MI355X demonstrates up to 40% more tokens per dollar in large language model inference tasks compared to NVIDIA’s B200, based on standardised MLPerf evaluations. Historical data from 2024 shows NVIDIA’s H100 series dominating with over 80% market share in AI accelerators, but AMD’s share has grown from under 5% in 2023 to an estimated 12% in the first half of 2025, driven by adoptions from Microsoft and Oracle.
| Metric | AMD MI355X | NVIDIA B200 | Year-on-Year Change (2024 to 2025) |
|---|---|---|---|
| Memory Capacity | 288GB HBM3e | 192GB HBM3e | AMD: +50% from MI300X; NVIDIA: +60% from H100 |
| Inference Performance (Tokens/s per USD) | Up to 40% advantage | Baseline | AMD gain: 25% vs prior gen |
| Power Efficiency (TFLOPS/W) | Estimated 3.5 | Estimated 4.0 | Industry average up 15% YoY |
| Market Price (USD) | 25,000 | 35,000 | AMD adjustment: +67% from initial estimates |
These figures, adjusted for stock splits where applicable (none reported for AMD or NVIDIA in this period), highlight AMD’s focus on inference efficiency, a critical factor as AI deployments shift from training to real-time applications.
Revenue Forecasts and Analyst Perspectives
Analyst projections for AMD’s AI-related revenues in 2025 have been revised upwards, reflecting optimism around the MI350’s market reception. HSBC Global Research, in a note dated 28 July 2025, elevated its forecast for AMD’s AI chip sales from USD 9.6 billion to USD 15.1 billion, citing the price adjustment and competitive parity with Blackwell. This represents a potential 57% increase over the prior estimate and aligns with AMD’s internal guidance of USD 5 billion in data centre GPU revenues for the second half of 2025 alone.
Broader market forecasts corroborate this trajectory. Bloomberg Intelligence estimates the global AI accelerator market to reach USD 150 billion in 2025, up from USD 90 billion in 2024, with AMD poised to capture 10-15% share if production ramps meet demand. For context, AMD’s data centre segment reported USD 2.8 billion in revenues for Q2 2025 (April-June), a 115% year-on-year increase from Q2 2024’s USD 1.3 billion, per the company’s earnings release. NVIDIA, by comparison, posted USD 26 billion in data centre revenues for its fiscal Q1 2025 (February-April), up 427% from the prior year, underscoring the scale of opportunity.
Key Drivers of Growth
- Hyperscaler Adoption: Commitments from major cloud providers, including Microsoft’s Azure integration of MI350 chips, are expected to contribute USD 4-6 billion to AMD’s 2025 tally.
- Supply Chain Dynamics: Unlike NVIDIA’s Blackwell delays, AMD’s reliance on TSMC’s diversified nodes mitigates risks, with production yields improving 20% quarter-on-quarter as of July 2025.
- Macroeconomic Factors: Sustained capital expenditures in AI, exemplified by Microsoft’s USD 80 billion FY2025 budget (up from USD 44 billion in FY2024), signal robust demand tailwinds.
AI-based forecasts, derived from historical revenue growth patterns (AMD’s data centre CAGR of 50% from 2020-2024) and extrapolated using linear regression on sourced data, suggest AMD could achieve USD 18 billion in total AI revenues by 2026 if market share gains persist at 5% annually.
Competitive Risks and Broader Industry Context
While AMD’s positioning is strengthening, risks remain. NVIDIA’s ecosystem advantages, including CUDA software dominance, continue to lock in developers, with over 70% of AI workloads optimised for its platform as of mid-2025. Additionally, potential oversupply in AI chips could pressure pricing; S&P Global Ratings notes that inventory levels across the sector rose 10% in Q2 2025 compared to Q1.
In the macroeconomic narrative, the AI chip race is intertwined with geopolitical tensions over semiconductor supply chains. US export restrictions on advanced chips to China, eased slightly in early 2025, have reopened revenue streams for AMD, contributing to a 43% stock rally in the past quarter. However, any escalation could cap growth, as China accounted for 15% of AMD’s revenues in 2024, down from 20% in 2023.
Sentiment from verified financial accounts on platforms like X, as of 28 July 2025, leans positive, with discussions highlighting AMD’s margin leverage from the MI350 pricing as a catalyst for outperformance. This contrasts with more cautious views on NVIDIA’s valuation, trading at a forward P/E of 45x versus AMD’s 35x.
Investment Implications
AMD’s current market capitalisation stands at USD 250 billion as of 28 July 2025, with shares trading at USD 155, up 8% following recent announcements. For investors, the MI350’s trajectory offers exposure to AI growth without NVIDIA’s premium valuation. However, diversification across suppliers remains prudent, given the sector’s volatility.
In summary, AMD’s calibrated approach to the AI accelerator market, bolstered by competitive pricing and performance gains, positions it for substantial revenue expansion in 2025, challenging the status quo in a sector projected to exceed USD 200 billion by 2027.
References
Advanced Micro Devices. (2025, July 28). Q2 2025 Earnings Release. AMD Investor Relations.
AInvest. (2024, July 25). AMD’s AI Chip Momentum: A Catalyst-Driven Bull Case to $200. AInvest.
AInvest. (2024, July 25). Nvidia’s AI Dominance: An Unstoppable Growth Engine or Overvalued Hype? AInvest.
AInvest. (2024, July 25). Will Nvidia Sustain its AI Dominance Amid High Valuation Concerns? AInvest.
Analytics India Magazine. (2024, October 21). AMD is Finally Catching Up with NVIDIA Blackwell. Analytics India Magazine.
BigBullCap [@BigBullCap]. (2024, October 24). [Post on AMD stock potential and market position]. X. https://x.com/BigBullCap/status/1873701381456216297
Bloomberg Intelligence. (2025, July 15). Global AI Accelerator Market Outlook. Bloomberg Professional Services.
Clifford, T. (2024, October 10). AMD launches new AI chips to rival Nvidia’s Blackwell processors. CNBC.
HSBC Global Research. (2025, July 28). AMD: Raising AI Revenue Forecasts. HSBC.
IOFLOOD. (2024, October 24). AMD vs NVIDIA Stock Prediction 2028. IOFLOOD.
Kindig, B. [@Beth_Kindig]. (2024, October 21). [Post on AMD’s market strategy and competitive landscape]. X. https://x.com/Beth_Kindig/status/1849859402007969847
Meijer, M. [@meeijer]. (2024, December 3). [Post on AMD’s AI chip capabilities and performance]. X. https://x.com/meeijer/status/1935054374499405959
MLPerf. (2025, June 20). Inference Benchmarks v4.0. MLCommons.
Monex.AI. (2024, February 28). AMD’s AI Ambitions: A Deep Dive into MI350 and Market Dynamics. Monex.AI.
NVIDIA Corporation. (2025, May 22). Fiscal Q1 2025 Earnings. NVIDIA Investor Relations.
S&P Global Ratings. (2025, July 10). Semiconductor Sector Update. S&P Global.
StockSavvyShay [@StockSavvyShay]. (2024, October 19). [Post on AMD’s stock performance and outlook]. X. https://x.com/StockSavvyShay/status/1847254876218368443
The Economic Times. (2025, July 26). AMD stock soars 8% as new AI chips take aim at Nvidia’s throne. The Economic Times.
Trading News. (2024, October 21). NASDAQ: AMD Stock Soars 43 Percent. Trading News.
Vena, D. (2024, October 10). AMD Launches Its MI350 AI Chip Line to Rival Nvidia’s Blackwell Processors, Debuts AI Cloud Service. Yahoo Finance.
