How did Meta turn MTIA from an inference chip into a gigawatt-scale program?
Meta began by improving an inference accelerator for ranking and recommendation, then tested in-house training silicon as it sought to reduce reliance on Nvidia and other suppliers. That did not eliminate outside capacity: Meta discussed and then signed a deal to rent Google TPUs for its next models. In parallel, Meta reported hundreds of thousands of deployed MTIAs, published four more generations on a six-month cadence, and committed an initial 1 GW of MTIA with Broadcom through 2029. It later publicly detailed MTIA 400 and introduced Tahan, an Ethernet scale-up switch purpose-built for MTIA 300 training accelerators.
Trace the chain
MTIA expansion
- 1MTIA 200 improves inferenceLaunch
Meta's second-generation MTIA doubles compute for ranking and recommendation models
Meta detailed its next MTIA inference chip, a year after revealing the first, with more than twice the compute and memory bandwidth of v1 and an SRAM-heavy design tuned for the ads ranking and recommendation models that run its feeds. Meta framed MTIA as a long-term program to own the most efficient silicon for its own workloads rather than a general challenger to GPUs.
Meta AI Blog ·
This new version of MTIA more than doubles the compute and memory bandwidth of our previous solution while maintaining our close tie-in to our workloads.
Why it matters here. The second MTIA's more than doubled compute and memory bandwidth gave Meta a larger in-house inference base from which it could explore work beyond ranking and recommendation.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
- 2Meta tests training siliconReport
Reuters: Meta is testing its first in-house training chip, a step beyond MTIA's inference role
Reuters, via TechCrunch, reported that Meta was testing an in-house chip for training AI systems as part of its strategy to reduce reliance on Nvidia and other hardware makers. It was the first sign that MTIA would move from recommendation inference into training, which Meta confirmed a year later with the MTIA 300 to 500 roadmap.
TechCrunch ·
Meta is reportedly testing an in-house chip for training AI systems, a part of a strategy to reduce its reliance on hardware makers like Nvidia.
Why it matters here. Meta's testing of an in-house training chip to reduce reliance on Nvidia and other suppliers showed that custom silicon was expanding beyond inference, even as Meta next considered renting and deploying Google TPUs for additional capacity.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
Gigawatt execution
- 6Four-generation MTIA roadmapAnnouncement
Meta lays out four MTIA generations in two years: 300, 400, 450 and 500, on a six-month cadence
Meta said it had deployed hundreds of thousands of MTIA chips, renamed the first two parts MTIA 100 and 200, and published a roadmap of four more generations through 2027 that widens the chips from recommendation inference to recommendation training and generative-AI inference, built with Broadcom on modular chiplets. The post committed Meta to iterating silicon at a consumer-hardware pace.
Meta AI Blog ·
Since introducing MTIA 100 and 200, we have accelerated MTIA development across four successive generations: MTIA 300, 400, 450, and 500. These new chips have either already been deployed or are scheduled for…
Why it matters here. With hundreds of thousands of MTIAs deployed and four further generations scheduled on a six-month cadence, Meta had an in-house program whose next step added an initial 1-GW deployment commitment.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
- 7Broadcom gigawatt commitmentAnnouncement
Meta commits to a gigawatt of MTIA with Broadcom through 2029; Hock Tan leaves Meta's board
Meta extended its Broadcom partnership through 2029 with an initial 1 GW deployment of MTIA and multiple gigawatts to follow, on what Broadcom said would be the first 2nm AI silicon. The same day Meta disclosed that Broadcom's CEO would not stand for re-election to its board after two years, moving to an advisory role on the chip program. Tan had told analysts in March that the MTIA roadmap was 'alive and well'.
CNBC ·
Meta and Broadcom on Tuesday announced a sweeping deal that extends an existing partnership between the two companies for the design of Meta's custom in-house AI accelerators through 2029. At the same time, Meta said…
Why it matters here. Meta's initial 1-GW MTIA commitment with Broadcom, with multiple gigawatts to follow through 2029, coincided with the subsequent public presentation of MTIA 400, a generation on its roadmap.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
- 8MTIA 400 is detailedReport
Hot Chips 2026: Meta and Microsoft open up MTIA 400 and Maia 200; AI accelerators become the biggest track
Meta presented the MTIA 400 from its March roadmap and Microsoft gave its deepest account of Maia 200's architecture: 140 billion transistors, six HBM3E stacks and a 750 W envelope, a deliberately less aggressive design than its rivals'. ServeTheHome noted that AI accelerators had overtaken CPUs as the largest track at the conference, a measure of how much engineering the hyperscalers now put into their own silicon.
ServeTheHome ·
Though by no means a small chip at 140 billion transistors, Microsoft has paired the accelerator with just 6 stacks of HBM3e memory, and the total chip TDP is just 750 Watts.
Why it matters here. Meta publicly presented MTIA 400, documenting one of the generations previously listed on its March MTIA roadmap.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
MTIA training system
- 9Tahan scales MTIA 300Launch
Meta introduces Tahan Ethernet scale-up switch for MTIA 300
Meta introduced Tahan, an Ethernet-based scale-up switch. Meta said Tahan was purpose-built for its MTIA 300 training accelerators.
Engineering at Meta ·
We’re introducing Tahan, Meta’s Ethernet-based scale-up switch — purpose-built for our MTIA 300 training accelerators.
Why it matters here. Tahan’s purpose-built scale-up role for MTIA 300 training accelerators shows that Meta’s expansion beyond inference included supporting system networking as well as successive chip generations.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
What pushes back
- 3Meta seeks Google TPUs
Meta in talks to put Google TPUs in its data centers from 2027; Nvidia falls 4 percent
The Information reported that Meta was negotiating to deploy TPUs in its own data centers in 2027 and to rent them from Google Cloud in 2026, and that Google was now promoting TPUs for installation in customers' facilities rather than only in its cloud. Alphabet rose 4 percent and Nvidia fell as much as 7 percent on the report; a Google spokesperson said demand for both TPUs and Nvidia GPUs was accelerating.
CNBC ·
On Monday, The Information reported that Meta is considering using Google's tensor processing units, or TPUs, in its data centers in 2027. Meta may also rent TPUs from Google's cloud unit next year, the publication…
What it pushes against. The discussions to rent TPUs in 2026 and deploy them in Meta data centers from 2027 push against an interpretation that its own chips had made external capacity unnecessary.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
- 4AMD GPU capacity deal
Meta signs for up to 6 gigawatts of AMD GPUs with a custom MI450-based part and a matching warrant
AMD and Meta agreed a multi-year, multi-generation partnership to deploy up to 6 gigawatts of Instinct GPUs. Shipments for the first gigawatt were expected to begin in the second half of 2026 using a custom GPU based on the MI450 architecture, and AMD issued Meta a performance-based warrant for up to 160 million shares.
AMD Investor Relations ·
AMD has issued Meta a performance-based warrant for up to 160 million shares of AMD common stock, structured to vest as specific milestones associated with Instinct GPU shipments are achieved.
What it pushes against. The agreement to deploy up to 6 GW of AMD Instinct GPUs pushes against an MTIA-only reading of Meta’s capacity strategy.
- 5TPU rental becomes a deal
Meta signs a multibillion-dollar deal to rent Google Cloud TPUs for its next models
Three months after the first report, The Information said Meta had agreed to a multibillion-dollar deal to rent TPUs from Google Cloud to train and run its next-generation models, alongside multibillion-dollar purchases of Nvidia's Vera Rubin and AMD's MI400 GPUs announced the same month. Meta was buying from every accelerator vendor at once while its own MTIA program continued.
SiliconANGLE ·
Meta Platforms Inc. has reportedly agreed to a multibillion-dollar deal to rent Google Cloud’s custom artificial intelligence chips, according to a report in The Information today. The report cited anonymous sources as…
What it pushes against. The multibillion-dollar TPU rental agreement for Meta's next models confirms that external compute remained part of the plan alongside MTIA.
Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
Changes to this chain
- Proposal #216 accepted (publish:hyperscaler-custom-accelerators): insert after 8: [cause] Tahan scales MTIA 300 -> hyperscaler-custom-accelerators#2145: Tahan’s purpose-built scale-up role for MTIA 300 training accelerators shows that Meta’s expansion beyond inference included supporting system networking as well as successive chip generations.; answer: Meta began by improving an inference accelerator for ranking and recommendation, then tested in-house training silicon as it sought to reduce reliance on Nvidia and other suppliers. That did not eliminate outside capacity: Meta discussed and then signed a deal to rent Google TPUs for its next models. In parallel, Meta reported hundreds of thousands of deployed MTIAs, published four more generations on a six-month cadence, and committed an initial 1 GW of MTIA with Broadcom through 2029. It later publicly detailed MTIA 400 and introduced Tahan, an Ethernet scale-up switch purpose-built for MTIA 300 training accelerators.
- Proposal #198 accepted (publish:hyperscaler-custom-accelerators): insert after 3: [counter] AMD GPU capacity deal -> amd-instinct-ai-gpus#108: The agreement to deploy up to 6 GW of AMD Instinct GPUs pushes against an MTIA-only reading of Meta’s capacity strategy.
- Proposal #143 accepted (verify:v2): why of step 5 -> With hundreds of thousands of MTIAs deployed and four further generations scheduled on a six-month cadence, Meta had an in-house program whose next step added an initial 1-GW deployment commitment.; why of step 6 -> Meta's initial 1-GW MTIA commitment with Broadcom, with multiple gigawatts to follow through 2029, coincided with the subsequent public presentation of MTIA 400, a generation on its roadmap.; why of step 7 -> Meta publicly presented MTIA 400, documenting one of the generations previously listed on its March MTIA roadmap.
- Proposal #39 accepted (verify:v1): why of step 2 -> Meta's testing of an in-house training chip to reduce reliance on Nvidia and other suppliers showed that custom silicon was expanding beyond inference, even as Meta next considered renting and deploying Google TPUs for additional capacity.; why of step 7 -> Meta publicly presented MTIA 400, one of the generations on its March roadmap, rather than leaving that roadmap undescribed.
Timelines in this chain
- Custom AI Silicon at the Hyperscalers: TPU, Trainium, Maia, MTIA and the OpenAI–Broadcom Chip
- AMD Instinct: From MI300X to Helios and the Gigawatt Deals
- HBM Supply Race: SK hynix, Samsung and Micron from HBM3E to HBM4E
- Nvidia Data Center GPUs: Hopper, Blackwell and Vera Rubin
A chain is written by the editor from entries already published on the timelines above, each of which was read against its source before publication. The chain changes when those entries do. Data and terms.