Enterprises prefer non-Nvidia chips for AI acceleration
Interest in Nvidia alternatives grows in corporate AI infrastructure.
1Nvidia's dominance challenged
Non-Nvidia chips lead by 14 points in enterprise AI accelerator evaluation lists. A July survey of 170 AI infrastructure professionals revealed that 39.4% plan to evaluate alternatives like AWS Trainium, Google TPU, and AMD Instinct in the next 12 months, compared to just 25.3% for Nvidia's next-gen GPUs. This shift indicates companies are no longer viewing Nvidia as the only viable option, building real optionality into their accelerator strategies.
2Azure shows significant growth
Microsoft Azure demonstrated the largest production adoption growth among major platforms. Azure usage jumped from 29% in June to 47.1% in July - an 18.1 percentage point increase. This surge partially reflects the survey's respondent base shifting toward larger enterprises (57% from organizations with 1,000+ employees vs. 37% in June). Azure adoption consistently correlates with company size, showing its strength in enterprise environments.
3Google Gemini maintains lead
Google's Gemini platform remained the most widely used in both survey waves. Production usage grew from 41.1% in June to 47.6% in July, slightly ahead of Azure. OpenAI's production adoption increased from 40.2% to 49.4%, while Anthropic's more than doubled from 12.1% to 24.7%. These figures demonstrate increasing diversity in corporate AI infrastructure choices beyond single-vendor solutions.
4GPU utilization efficiency improves
Companies running their own GPUs showed marked efficiency gains. Those using less than half capacity dropped from 83% to 69%, while operations above 50% utilization rose from 13% to 23%. Infrastructure effectiveness metrics also evolved operationally - respondents prioritizing uptime/reliability increased from 42.1% to 51.2%, and those focusing on throughput grew from 21.5% to 24.7%.
5Platform change urgency declines
Plans for immediate platform changes decreased significantly. Respondents expecting changes within 0-3 months dropped 9.5 points to 28.8%. Those planning changes in 3-6 months rose 4.1 points, while 6-12 month plans increased 5.3 points. Approximately 40% maintain no change plans. This shift reflects companies prioritizing optimization of existing infrastructure over platform replacements.
6Selection criteria evolve
Integration with existing cloud/data stacks remained the top selection factor at 40%. Performance prioritization jumped from 24.3% to 35.3%, while cost-per-million-tokens consideration doubled from 7.5% to 15.9%. GPU access importance rose from 18.7% to 23.5%. Conversely, broad TCO focus declined from 34.6% to 21.8%, indicating a market shift toward workload-specific evaluation over general infrastructure planning.
7C-suite interest in alternatives
C-level executives showing interest in non-Nvidia accelerators increased from 42.9% to 57.1%. Final decision-makers considering alternatives grew from 35.4% to 50%. The trend was most pronounced in SMBs: 251-1,000 employee firms rose from 41.4% to 53.2%, while 101-250 employee companies surged from 33.3% to 57.7%. This signals accelerator diversity becoming a strategic infrastructure consideration beyond just engineering teams.
8Infrastructure effectiveness metrics
Uptime/reliability as key effectiveness measures grew from 42.1% to 51.2%. Throughput focus increased from 21.5% to 24.7%. Implementation ease improved from 3.84 to 4.04 on a 5-point scale. Overall satisfaction edged up slightly from 4.07 to 4.14, while perceived value remained steady at ~3.9. These metrics suggest enterprises are raising their infrastructure expectations even as capabilities improve.
9SMB adoption trends
Small and medium businesses showed particularly strong interest in Nvidia alternatives. Evaluation likelihood grew from 41.4% to 53.2% in 251-1,000 employee firms, and from 33.3% to 57.7% in 101-250 employee companies. This reflects SMBs' cost sensitivity and strategic approach to accelerator selection. The data confirms that hardware diversity has transitioned from technical consideration to strategic infrastructure decision across organization sizes.
10New era in AI infrastructure
The survey points to a new phase in AI infrastructure development. Enterprises are actively building optionality rather than relying on single vendors, particularly at executive decision-making levels. Evolving effectiveness metrics reveal growing demands for reliability and efficiency. This diversification trend could foster more competitive dynamics in the AI hardware ecosystem, benefiting organizations through improved choice and potentially better pricing models.
Original article: VentureBeat (feeds.feedburner.com) — Şirketler Nvidia dışı çipleri tercih ediyor