Не только Китай: Jalapeño OpenAI и глобальная волна custom chips
Custom AI silicon — уже глобальный феномен. TrendForce (2026): shipments hyperscaler custom chips растут на 44,6%, general-purpose GPUs — на 16,1%. Впервые custom silicon обгоняет по growth rate.
| Компания | Chip project | Стадия | Workload | Key signal |
|---|---|---|---|---|
| DeepSeek | Unnamed inference ASIC | Early R&D | Inference | $7,4B funding; private hiring; unconfirmed |
| Alibaba (T-Head) | Zhenwu 810E / M890 | Mass production | Train + infer | 560K+ shipped; ~$1,4B annual revenue |
| Huawei | Ascend 950 series | Mass production | Train + infer | DeepSeek V4 adapted; orders surging |
| OpenAI | Jalapeño (Broadcom) | Tape-out done | Inference | 9-month design cycle; deploy late 2026 |
| TPU v6/v7 | At scale | Train + infer | Gemini runs end-to-end on TPU | |
| Amazon | Trainium3 / Inferentia | Commercial | Both | Anthropic uses Trainium at scale |
| Microsoft | Maia 100 | Deploying | Inference | Powers Azure / OpenAI workloads |
| Meta | MTIA | Internal | Inference | Recommendations; once scrapped and rebuilt |
| Anthropic | Samsung custom chip talks | Exploring | TBD | The Information, July 2026 |
| Zhipu AI | Evaluating custom chip | Early | Inference | The Information, July 2026 |
Key dates: 24 июня OpenAI Jalapeño; 2 июля Anthropic–Samsung 2nm talks; 7 июля Reuters on DeepSeek; 7 июля The Information on Zhipu. См. также deep dive по Jalapeño.
Что Reuters реально написал (и что DeepSeek не подтвердил)
Bottom line: можно писать «Reuters и другие сообщают, что DeepSeek запустил custom inference chip program». Нельзя писать «Liang Wenfeng officially announced chip development». Тегайте: sources familiar / early stage / unconfirmed.
30-second summary: скорее всего real, но early. Нет CEO announcement. T-Head уже в mass production. Economics drives shift; geopolitics accelerates.
Inference-only ASIC: заточен под serving, не под training clusters.
Старт ~mid-2025: описано как «about a year ago»; всё ещё early stage.
Supply chain talks: переговоры с chip designers, foundries, memory vendors.
Quiet hiring: chip engineers recruited privately, не на public job boards.
Dual-dependency play: снижение reliance на Nvidia и Huawei Ascend — DeepSeek уже runs on Ascend.
| Credibility factor | Assessment |
|---|---|
| Source tier | High. Reuters standard «three people familiar with the matter» |
| Official confirmation | None as of this writing |
| Circumstantial evidence | Strong. ~$7,4B (~51B RMB) June 2026 round earmarked for chips and domestic compute; IDC hiring; UE8M0 FP8 format — hardware-software co-design signal |
| Contradictory takes | Partnership и in-house R&D run in parallel — Ascend live, custom silicon early |
Что CEO DeepSeek Liang Wenfeng говорил про chips и compute
Liang Wenfeng дал мало публичных интервью. Самый ценный source — два deep dive с Waves (暗涌) в мае 2023 и июле 2024. Он никогда не анонсировал chip program, но обозначил strategic motive.
«Our real challenge has never been capital — it is export controls on advanced chips.» — Liang Wenfeng, Waves interview, July 2024
4× compute overhead: domestic training efficiency lags ~1×, data efficiency ещё ~1× — ~4× total compute для parity.
Missing tech community: domestic chips lack first-hand developer community; кто-то должен стоять на frontier.
Endless compute hunger: researchers always want more capacity; DeepSeek deploys as much compute as it can.
Founder remarks ≠ product launch: Reuters описывает company actions (hiring, vendor talks), не CEO announcement.
Co-design signals: UE8M0 FP8 и MLA architecture optimizations point toward hardware-specific tuning.
T-Head Alibaba уже шипит — ставка Jack Ma 2018 окупается в 2026
Не пишите «Jack Ma недавно сказал, что Alibaba будет делать chips». Точная дуга: Jack Ma задал T-Head strategy в 2018, Joe Tsai объяснил export-control pressure в 2024, CEO Wu Yongming раскрыл mass-production numbers в 2026.
| Figure | Role | Public chip-related stance |
|---|---|---|
| Jack Ma | 2018 strategic sponsor | Named T-Head, elevated chips to group strategy |
| Joe Tsai | Chairman | 2024 podcast: US export limits hit Alibaba Cloud; long-term faith in domestic semiconductors |
| Wu Yongming | CEO | 2026 earnings call: 470K+ T-Head AI chips delivered; billion-yuan annual revenue; IPO possible |
| Model | Timeline | Highlights |
|---|---|---|
| Hanguang 800 | 2019 | Early AI inference chip |
| Zhenwu 810E | Jan 2026 | Train + infer; 96GB HBM2e; between Nvidia A800 and H20; in production |
| Zhenwu M890 | 2026 | 144GB memory, 800GB/s interconnect, ~3× 810E |
| Zhenwu V900 | Planned Q3 2027 | 216GB, 1200GB/s interconnect |
| Zhenwu J900 | Planned Q3 2028 | Next-gen parallel compute architecture |
WSJ: новые Alibaba chips support Nvidia CUDA ecosystem, easing engineer migration (unlike Huawei path). Manufacturing shifted from TSMC toward domestic foundries (industry points to SMIC 7nm-class flows).
560 000+ units shipped (H1 2026).
~$1,4B annual revenue; 400+ enterprise customers on Zhenwu clusters.
T-Head registered capital raised to ~$140M (Jun 2026); Alibaba pledged ~$52B over three years for cloud and AI infrastructure.
Почему tech giants строят custom AI chips: cost, control и «Nvidia tax»
One-liner: AI competition сдвинулась от «who has the best model» к «who has the cheapest, most controllable compute».
Economics — inference is the rent: training — down payment, inference — monthly rent. At ChatGPT-scale DAU inference spend exceeds training. Custom ASICs cut TCO 30–65% at scale; per-token costs down 30–40%. Nvidia datacenter GPU gross margins exceed 70% — in-house silicon converts permanent «GPU tax» into one-time R&D.
Supply chain resilience: US export controls, allocation shortages, single-vendor risk — не только «national security», но и predictable supply.
Hardware-software co-design: general GPUs trade efficiency for flexibility; ASICs — наоборот для known workloads. Jalapeño targets real ChatGPT serving (KV cache, batching, latency).
Bargaining power: даже partial self-supply strengthens Nvidia negotiations и enables full-stack «model + cloud + chip» story.
Energy: inference ASICs optimize performance-per-watt — critical at gigawatt-scale datacenters.
| Dimension | Training | Inference |
|---|---|---|
| Workload | Dynamic, experimental, architecture churn | Static model, predictable request patterns |
| Software moat | CUDA stack (cuDNN, NCCL, Nsight) | Hand-tuned kernels for fixed models |
| Chip priority | Peak FLOPS + programmability | Throughput, latency, cost per token |
| Economics | Large one-time capex | 24/7 at scale — bigger ongoing spend |
| Verdict | Training stays Nvidia territory; inference — custom ASIC battleground. | |
Morgan Stanley (via Reuters Breakingviews): 24 000-GPU Blackwell cluster costs ~$852M in hardware; equivalent Google TPU cluster ~$99M.
Six-step decision guide:
Separate rumor from announcement: пишите «reportedly» until DeepSeek confirms.
Split training vs inference planning: frontier training still needs Nvidia; inference — where ASICs win.
Track parallel paths: DeepSeek on Ascend is live; custom silicon is early.
Model TCO, not sticker price: focus on per-token cost и multi-year capex ROI.
Price early-project risk: Meta scrapped and rebuilt MTIA; architecture shifts can obsolete ASIC designs.
Decouple local agents from cloud inference: chip economics mostly affect API pricing; local Cursor/Codex agents и iOS CI still need stable macOS hosts 24/7.
Disclaimer: DeepSeek не подтвердил chip project officially as of this writing. Sources: Reuters, WSJ, OpenAI official blog, Waves interviews, Alibaba filings. Not investment advice.
Reality check: API-only local agents — linear token spend и model availability risk under export controls; personal Macs for mixed training and agents hit unified-memory limits и sleep interruptions; macOS VMs break EULA и restrict Xcode signing. Для iOS CI/CD, local LLM inference и AI agent automation in production KVMNODE dedicated Mac Mini M4 cloud rental — usually the better fit: Apple Silicon unified memory for Metal inference, 24/7 uptime, flexible daily/weekly/monthly billing. См. цены, оформить заказ, центр помощи.
Последнее обновление: 10 июля 2026 · Sources: Reuters, OpenAI official, WSJ, Caixin Global, Waves interviews, Alibaba/T-Head public disclosures