DEPLOY

AI chip · Meta

Meta MTIA v2

Second-generation MTIA, ~3x compute per package, doubled memory bandwidth. Announced April 2024.

Deployed in 2 named data centers.

Market position

Meta's second-generation in-house silicon for recommendation and ranking (news feed, Reels, ads). Not sold. Deployed at scale across every Meta AI campus (Prometheus, Hyperion) alongside NVIDIA. Signals Meta's long-term aim to reduce NVIDIA dependency for inference.

What fits in 128 GB

Which open-source LLMs run on one Meta MTIA v2, by precision. Weights only: add roughly 15% headroom for real serving. If a model does not fit at FP16, try INT8 or INT4 (smaller quality trade-off than most people expect).

ModelParamsFP16INT8INT4
Llama 3.1 8B
dense
8 B16 GB ✓8 GB ✓4 GB ✓
Llama 3.1 70B
dense
70 B140 GB ✗70 GB ✓35 GB ✓
Llama 3.1 405B
dense
405 B810 GB ✗405 GB ✗203 GB ✗
Llama 3.3 70B
dense
70 B140 GB ✗70 GB ✓35 GB ✓
DeepSeek V3
MoE (671B total, 37B active per token)
671 B1342 GB ✗671 GB ✗336 GB ✗
DeepSeek R1
MoE (671B total, 37B active per token)
671 B1342 GB ✗671 GB ✗336 GB ✗
Qwen 2.5 7B
dense
7 B14 GB ✓7 GB ✓4 GB ✓
Qwen 2.5 72B
dense
72 B144 GB ✗72 GB ✓36 GB ✓
Mixtral 8x7B
MoE (46.7B total, 12.9B active per token)
46.7 B93 GB ✓47 GB ✓23 GB ✓
Mixtral 8x22B
MoE (141B total, 39B active per token)
141 B282 GB ✗141 GB ✗71 GB ✓
Gemma 2 27B
dense
27 B54 GB ✓27 GB ✓14 GB ✓
Command R+
dense
104 B208 GB ✗104 GB ✓52 GB ✓
Kimi K2
MoE (1T total, 32B active per token)
1000 B2000 GB ✗1000 GB ✗500 GB ✗

Math: FP16 = params × 2 bytes; INT8 = params × 1 byte; INT4 = params × 0.5 bytes. MoE models sum every expert (full weights on disk), not the per-token active subset.

Common questions

How much does Meta MTIA v2 cost?

Meta MTIA v2 doesn't have a public launch list price. Vendors like this one usually price through direct sales rather than a public sheet. See list price per TFLOP for chips that do publish.

How much memory does Meta MTIA v2 have?

128 GB of LPDDR5, running at 205 GB/s. Straight from the vendor datasheet. See chips with the most memory for context.

How much power does one Meta MTIA v2 draw?

90 W at the chip. A full server draws more once you add CPU, memory, networking and cooling: see the rack power number above. Compare to other chips on perf-per-watt.

Which data centers use Meta MTIA v2?

2 named data centers run them, including Meta Hyperion and Meta Prometheus. See who has the most Meta MTIA v2 for the ranked list.

Which open-source LLMs fit on one Meta MTIA v2?

Of 13 open-source LLMs we track, 4 fit at FP16, 8 at INT8, and 9 at INT4 (for example Llama 3.1 8B, Llama 3.1 70B and Llama 3.3 70B). Full table above with each model's memory need.

See every answer we publish →

Key facts

Class
Compute SoC (AI accelerator)
Designer
Meta
Safety-critical?
No (data-center inference / training)
Record as of
2026-09-27
Most recent source
2024-04-10 (across 2 sources on this page)
Specifications (8 fields, click to expand)

Straight from the vendor datasheet. Dense throughput shown first; sparse (2:4) numbers in parentheses where the vendor publishes them. Full datasheet linked below.

Process node
TSMC N5
TDP
90 W
Memory
128 GB LPDDR5
Memory bandwidth
205 GB/s
INT8 (dense)
708 TOPS
Form factor
PCIe (internal Meta)
Announced
2024-04-10
Released
2024-05-01

Source: vendor datasheet

Generation

Foundry & process

Fabbed at
TSMC (all chips TSMC makes →)
Process node
TSMC N5

Compare with

See every chip comparison →

Data centers running Meta MTIA v2

Who has the most? →

Named data-center campuses with Meta MTIA v2 on site. Counts shown where the operator has published them; other rows are described in general terms.

  1. under constructionreported

    Meta's own AI accelerator generation deployed at Hyperion

    Meta AI

  2. partially energizedreported

    Meta's own AI accelerator generation deployed alongside Nvidia at Prometheus

    Meta AI

Sources

Adoption rows appear as we confirm each chip-in-robot pairing from a public source. See every chip we track for the full catalog or the Meta page.