AI and machine learning market to hit $1.2 trillion by 2035
Analyst firm IndexBox projects the global AI and machine learning market will reach $1.2 trillion by 2035, driven by a 22.5% CAGR. Growth is largely attributed to the shift toward custom ASICs and neuromorphic chips for edge inference, which are significantly more cost-efficient than general-purpose GPUs for specific industrial and streaming applications.
Key Takeaways
- Edge AI inference devices are growing at 25-30% annually, outpacing decentralized cloud demand.
- Custom-designed ASICs and neuromorphic chips reduce per-operation costs by 40-60% versus general-purpose GPUs.
- Hardware remains the dominant value layer, accounting for over 55% of total market revenue.
- Foundry concentration persists, with the top three facilities controlling 90% of leading-edge AI chip production.
Why It Matters
For streaming video, this forecast signals a transition from cloud-heavy processing to decentralized, cost-optimized edge hardware. The move toward ASICs suggests that the historical reliance on high-cost NVidia hardware for video encoding and recommendation engines may subside as purpose-built silicon delivers better performance-per-watt. In the broader ecosystem, this shifts the competitive advantage from those with the largest cloud budgets to those with the most efficient proprietary hardware stacks. Operators should watch for a 30% jump in custom silicon inference market share by 2030, which will likely commoditize standard AI models while raising the barrier for breakthrough hardware integration.
Additional Context
The push for custom silicon is already materializing among streaming and cloud giants. Per Reuters and Seeking Alpha (August 2025), Broadcom currently holds an estimated 75% share of the custom AI ASIC market, largely through its work with Google, Meta, and ByteDance. Marvell follows with approximately 5%, securing major design wins with AWS for the Trainium and Inferentia series and with Microsoft for the Maia chips. These purpose-built accelerators are becoming essential as inference costs balloon; for instance, OpenAI reportedly faced a $20.9 billion operating loss in 2025 due to massive inference expenses, as reported by SemiAnalysis (July 2026). Efficiency gains are the primary motivator for this hardware shift. Technical benchmarks from early 2026 show that while general-purpose GPU clusters in production environments often see only 5% to 30% utilization for inference tasks, dedicated ASICs like Google’s TPU can achieve rates between 80% and 90%. This translates to significant operational savings; per industry data from July 2026, shifting inference workloads from GPUs to specialized chips reduced monthly compute costs by 65% for high-volume AI applications like image generation. This trend is forcing a diversification of the supply chain, as companies seek to avoid the 71% gross margins commanded by dominant merchant silicon vendors. At the edge, on-device AI is moving from experimental to mandatory. Per MarketIntelo (May 2026), the edge AI inference chip market reached $9.5 billion in 2025 and is projected to hit $57.8 billion by 2034. This growth is supported by recent regulatory mandates, such as the EU’s Data Act (effective September 2025), which requires sensitive personal and operational data to be processed locally. For streaming providers, this means the next generation of smart TVs and set-top boxes, powered by NPUs from Qualcomm and MediaTek, will likely handle more complex personalization and upscaling tasks locally rather than relying on back-end server calls.
Read full article at indexbox.io
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source