JPMorgan says Microsoft Maia 300 inference performance improves 40%, and million-chip deployment may reduce costs
BIBIBI
AT A GLANCE
The research report states that Microsoft Maia 300 has improved energy efficiency and throughput, and Microsoft plans to 2027 deploy more than 1000000 chips.
Article
JPMorgan: Microsoft Maia300 inference performance improves 40%, and million-chip deployment reduces costs
BlockBeats news,October 14, JPMorgan October 12, 2026 research report states that Microsoft Maia 300 has higher tokens per megawatt than leading GPU systems by 40%, and under a full 96 accelerator configuration, rack-level token throughput is Maia 200's 7 times; Maia 200 has higher tokens per watt than GPU infrastructure by 1.7 times; Reuters reported that Microsoft plans to 2027 deploy more than 1000000 chips in Maia 300; Microsoft 2027 fiscal-year capital expenditures are expected to be approximately 210000000000 USD，2026 in the calendar year approximately 175000000000 USD.
JPMorgan believes that Maia becoming the primary inference system will improve Copilot performance and reduce cost per compute unit. Microsoft cloud gross margin is expected to improve; it maintains an Overweight rating and a price target of 625 USD.
Original link https://m.theblockbeats.info/flash/371713
Key points
01
JPMorgan October 12, 2026 research report states that Microsoft Maia 300 has higher tokens per megawatt than leading GPU systems by 40%。
02
Under a full 96 accelerator configuration,Maia 300 rack-level token throughput is Maia 200's 7 times.
03
Maia 200 Tokens per watt are higher than GPU infrastructure by 1.7 times.
04
Reuters reported that Microsoft plans to 2027 deploy more than 1000000 Maia 300 chips.
05
Microsoft 2027 fiscal-year capital expenditures are expected to be approximately 210000000000 USD，2026 in the calendar year approximately 175000000000 USD。
06
JPMorgan believes that Maia becoming the primary inference system will improve Copilot performance and reduce cost per compute unit, potentially improving Microsoft cloud gross margin, while maintaining an Overweight rating and a price target of 625 USD。
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
This news item relays a JPMorgan research report and Reuters reporting. Its core focus is the inference efficiency, deployment scale, and expected capital expenditures of Microsoft's in-house Maia chips. If the relevant data and plans materialize, they may reduce AI compute costs and improve cloud business profits, but the information has not been directly confirmed by Microsoft or the original reports.
Why it matters to readers
Maia 300's performance and large-scale deployment plans may affect the market's assessment of Microsoft AI infrastructure costs, Copilot performance, and cloud gross margin.
It can be understood as: Microsoft hopes to use its in-house chips to run AI and reduce compute costs through higher energy efficiency per unit.
Risks and unknowns
The information comes from BlockBeats' relay of a JPMorgan research report and Reuters reporting.
The evidence does not provide direct confirmation from Microsoft or the original reports regarding the relevant data and deployment plans.
Maia 300 Whether the stated performance and cost advantages will be realized has not been confirmed by the evidence.
Whether Microsoft has officially confirmed 2027 deployment of more than 1000000 Maia 300 chips has not been confirmed by the evidence.
Related Developments
Loading event timeline…
Related concepts
Inference performance
This term is not in the glossary yet. Browse related concepts in the glossary.
