상태 표시 — 발표된 목표치는 '발표된 목표'로 표기됩니다
주목 목록
주목해야 할 대상: 도전자 실리콘, 차세대 플랫폼, 새로운 패브릭과 포토닉스, 전력 파고, 그리고 이 데이터베이스 전체를 재편할 수 있는 아키텍처 베팅.
Challenger silicon
8Shipping or credible alternatives to the incumbent GPUs — each betting on a different corner of the workload.
CW
Cerebras WSE-3Wafer-scale compute with tokens-per-second records — and an IPO on file.
GL
Groq LPUDeterministic inference at famous speed; Saudi-scale deployment deals.
ES
Etched SohuThe all-in transformer ASIC — if it ships, the specialization era begins.
DM
d-Matrix CorsairIn-memory compute aimed at cheap interactive serving.
FR
FuriosaAI RNGD150 W LLM inference; turned down Meta to stay independent.
RR
Rebellions REBELKorea's merged champion, fabbed by Samsung.
TB
Tenstorrent BlackholeJim Keller's open RISC-V stack, sold and licensed.
QC
Qualcomm Cloud AI / AI200Mobile-silicon economics re-entering the rack in 2026.
The next platform generation
5Already announced, already contracted — the hardware the 2026–27 buildout will install.
NR
NVIDIA RubinHBM4, the Vera CPU, NVL144 racks — the incumbent's next lap.
NR
NVIDIA Rubin CPXA cheaper GDDR7 chip just for long-context prefill — inference splits in two.
AI
AMD Instinct MI400 / MI450432 GB of HBM4 and the OpenAI 6 GW deal riding on it.
AT
AWS Trainium33 nm and a claimed 4x — the custom-silicon bet compounds.
GT
Google TPU v7 (Ironwood)9,216-chip pods for the inference age, with Anthropic as anchor.
New fabrics, memory and light
7The interconnect and memory transitions that decide whether open ecosystems can match proprietary racks.
UA
UALinkThe open answer to NVLink — 1,024 accelerators per pod, everyone-but-NVIDIA.
UE
Ultra EthernetEthernet rebuilt for AI; spec 1.0 shipped, silicon arriving.
CP
Co-Packaged OpticsLight moves onto the switch package; ~3x interconnect efficiency.
LI
LightmatterPhotonic interposers — light inside the package itself.
CA
Celestial AIOptical links to disaggregated memory — an exit from the HBM wall.
HB
HBM42,048-bit interfaces and foundry logic dies; memory and logic start merging.
CX
CXLPooled memory over PCIe wiring — slow burn, big if it lands.
New power
7The generation wave AI money is financing — targets, not certainties, and labeled as such.
SM
Small Modular Reactors (SMRs)Factory-built reactors sized for campuses; every date is a target.
OK
OkloFast reactors with a 12 GW Switch framework and a build-own-operate model.
XE
X-energyAmazon-backed pebble-bed reactors targeting ~2030.
KH
Kairos Hermes / Google FleetGoogle's 500 MW fleet order — first Gen IV construction permits in the US.
TN
TerraPower Natrium (Kemmerer)The furthest-along advanced reactor build in America.
FC
Fuel CellsBloom-style bridge power while the grid queue clears.
HC
Homer City Energy Campus4.5 GW of gas-for-AI on a demolished coal site — the biggest such plan.
Beyond the transformer
5Architecture bets that would redraw the hardware map if they win.
SS
State-Space Models (Mamba)Linear-time sequence models — production hybrids already ship.
WM
World ModelsLearning reality's dynamics, not just text — DeepMind, NVIDIA and World Labs all in.
M(
Marble (World Labs)Fei-Fei Li's spatial-intelligence bet, now a product.
G3
Genie 3Playable generated worlds — training grounds for agents and robots.
TT
Test-time Compute & ReasoningThe scaling axis that already rewrote inference demand once.
운영 원칙: 링크된 레코드의 모든 날짜는 제조사가 명시한 목표 일정이며, 그렇게 표시되어 있다. 이 페이지는 주장을 추적할 뿐 — 이를 보증하지 않으며, 투자 조언이 아니다.