Get all your news in one place.

100's of premium titles.
One app.

Get all your news in one place.

100's of premium titles. One news app.

FORGOTTEN HISTORY: Semianalysis | History Defined

OpenAI’s 700W Jalapeño ASIC outpaces 1,400W Nvidia flagship GPU — claims up to 1.9x throughput per kilowatt and 3.6x lower latency, co-developed with Broadcom

OpenAI arrived at Hot Chips on Tuesday with benchmarks claiming its first in-house chip beats Nvidia's GB300.

Tom’s Hardware

Google reportedly taps AMD to design next-generation TPU — hybrid AI ASIC could integrate on-package CPU cores for reinforcement learning

Google reportedly taps AMD to design next-generation TPU — hybrid AI ASIC could integrate on-package CPU cores for reinforcement learning

Google may be building a TPU with on-package CPU cores specifically for agentic and reinforced learning workloads, according to a rumor.

Tom’s Hardware

SpaceX Could Be Making $235 Billion From a Business It's Not Known For. SPCX Stock Investors Should Pay Attention.

SpaceX Could Be Making $235 Billion From a Business It's Not Known For. SPCX Stock Investors Should Pay Attention.

SemiAnalysis sees SpaceX's AI business reaching $235 billion in annual revenue by 2027, but at a huge infrastructure cost.

From analysis to the latest developments in health,
read the most diverse news in one place.

inkl

Elon Musk Says AI Will Be SpaceX's Biggest Business by September: Does the Math Check Out?

SpaceX (NASDAQ:SPCX) CEO Elon Musk says artificial intelligence will overtake everything else the company sells as soon as next month.

A machine made by China sends shockwaves in the chip industry

China has begun mass-producing its own lithography machines, challenging ASML's global dominance. This development marks a significant step towards Beijing's technological self-sufficiency goals. The move comes after years of US-led export controls on advanced chip technology. This breakthrough represents a milestone in China's semiconductor manufacturing capabilities.

The Economic Times

Nvidia Is No Longer The World's Most Valuable Company As The Chipmaker Rout Continues. Apple Has Taken The Lead

Nvidia stopped being the world's most valuable company as its stock continues to slide amid a broader rout in chipmakers.

International Business Times

Chinese memory maker CXMT posts blistering 466% leap in Shanghai IPO — bulk of spending to be focused on DRAM production, no HBM in sight

The prospectus splits 29.5 billion yuan across three projects.

Tom’s Hardware

inkl

Read news from The Economist, FT, and more,
with one subscription

Meta to use custom AMD Instinct MI400 accelerators with 144GB of HBM4 for select workloads, report claims — could dramatically reduce cost at the expense of versatility

Meta to use custom AMD Instinct MI400 accelerators with 144GB of HBM4 for select workloads, report claims — could dramatically reduce cost at the expense of versatility

Meta will reportedly use a custom version of AMD's Instinct MI400-series accelerators with a memory system cut to 144GB of HBM4, allegedly for select workloads only.

Tom’s Hardware

SMIC's third-gen 7nm node shows smaller metal pitch than Intel 18A, higher transistor density than TSMC N6 without EUV — analysis of N+3 shows significant advancement for Chinese semi manufacturing

SMIC's third-gen 7nm node shows smaller metal pitch than Intel 18A, higher transistor density than TSMC N6 without EUV — analysis of N+3 shows significant advancement for Chinese semi manufacturing

SMIC's N+3 process technology can achieve transistor density comparable to TSMC's N6 without using EUV lithography, but it fails to deliver performance or efficiency of modern production nodes.

Tom’s Hardware

Nvidia's Huang vows to deliver 'giant amounts' of Vera Rubin — company says that 'our roadmap is intact'

Nvidia's Huang vows to deliver 'giant amounts' of Vera Rubin — company says that 'our roadmap is intact'

Chief executive of Nvidia says the company is on track to produce 'giant amounts' of Vera Rubin-based machines, but fails to address rumored delays of Kyber NVL144 racks from 2027 to 2028.

Tom’s Hardware

A New Report Says Meta Platforms Could Overtake Google AI. How to Play META Stock Here.

A New Report Says Meta Platforms Could Overtake Google AI. How to Play META Stock Here.

Meta’s shares gained after SemiAnalysis predicted the company could leapfrog Google’s AI capabilities within months. How should investors position themselves in META stock?

Nebius Stock Sold Off on Meta’s Data Center News. Buy the Dip.

Nebius Stock Sold Off on Meta’s Data Center News. Buy the Dip.

Nebius stock sank on Meta’s cloud ambitions. SemiAnalysis believes Meta's AI expansion strengthens the buy-the-dip case.

Nvidia's Next AI Rack Hit A Manufacturing Snag. The Company's Fastest Product Cycle Just Slowed Down.

Nvidia's Next AI Rack Hit A Manufacturing Snag. The Company's Fastest Product Cycle Just Slowed Down.

Nvidia's next-generation Kyber NVL144 rack-scale AI system has reportedly been delayed by more than a year after manufacturing challenges emerged with a critical component.

International Business Times

Nvidia's Kyber rack for Rubin Ultra reportedly delayed to 2028, stopgap solution also axed due to customer pushback — Analyst firm SemiAnalysis says PCB midplane problems led to the delay [Updated]

Nvidia's Kyber rack for Rubin Ultra reportedly delayed to 2028, stopgap solution also axed due to customer pushback — Analyst firm SemiAnalysis says PCB midplane problems led to the delay [Updated]

Nvidia reportedly won't ship its Kyber NVL144 rack until 2028, a delay of more than 12 months.

Tom’s Hardware

Get all your
news in one place.

100's of premium titles.
One news app.

International Business Times UK

Meta Employees Burned $221M in AI Tokens in One Month: Now Big Tech Starts Counting Every AI Prompt

Companies are moving from unlimited AI usage to budgeted management, focusing on cost-effective productivity while maintaining significant gains. AI adoption continues to grow, with refined usage strategies.

Tom’s Hardware

Nvidia reportedly cancels quad-die Rubin Ultra GPU in favor of dual-GPU design, report claims — complex design purportedly scrapped over 'manufacturing execution concerns'

Nvidia reportedly abandons quad-dire Rubin Ultra GPUs in favor of dual-die Rubin Ultra due to 'manufacturing execution concerns.'

'The math doesn't work': Why your $200 AI subscription is secretly worth thousands

New research suggests heavy ChatGPT and Claude users are getting an enormous discount on compute. The catch: the math doesn't work, and the first cracks are already showing.

Tom’s Hardware

Chinese fab SMIC's 7nm metal pitch beats Intel 18A but lags 38% on density, teardown finds — Huawei's sanctions-beating HiSilicon Kirin 9030 is the first subject of SemiAnalysis's new teardown lab

SemiAnalysis has published the first teardown from its new in-house lab, focusing on the minimum local metal pitch on SMIC’s third-gen 7nm at 32.5nm.

Get all your
news in one place.

100's of premium titles.
One news app.

Why Experts Think Intel Should Raise Capital Right Now, and What It Means for INTC Stock

After a stunning AI-fueled comeback, Intel may have a chance to raise capital and supercharge its turnaround. But should shareholders be fine with the dilution?

Tom’s Hardware

AI costs spike as subscriptions hit pricing wall — firms turn towards Chinese LLMs, open-source models to extend budget

Companies look for cheaper alternatives as token costs for frontier AI models skyrocket, potentially impacting OpenAI and Anthropic's bottom lines. Subscriptions also take a bite out of these startup's profitability, as utilization rates higher than 5.7% could lead to losses.

The Economic Times

China's control over indium phosphide exports threatens AI data centre rollout

Anderson's visit was partly to raise the issue of delays in China's export licenses involving the highly strategic material, essential in manufacturing high-speed optical chips for AI data centres, said three sources familiar with the matter.

Tom’s Hardware

Memory will consume 30% of hyperscaler AI data center spending this year, a 4X increase over 2023 — Nvidia gets preferential supply terms well below standard market rates, says analyst firm

SemiAnalysis estimates on Thursday that memory will account for roughly 30% of total hyperscaler capex in calendar year 2026.

Tom’s Hardware

AMD denies report of MI455X delays as Nvidia VR200 systems are rumored to arrive early — company says Helios systems 'on target for 2H 2026'

AMD's Helios rack-scale solution based on the MI455X accelerators, the company's major hope for AI market, may slip to 2027, whereas Nvidia may speed up roll out of the Vera Rubin platform if the latest market rumors are to be believed.

Tom’s Hardware

Elon Musk says xAI will have more AI compute than everyone else combined within five years — Macrohard-branded Colossus 2 data center a nod to Musk's AI project to challenge Microsoft

Elon Musk claimed on X that xAI will have more computing power than everyone else combined in less than five years.

Tom’s Hardware

InferenceMax AI benchmark tests software stacks, efficiency, and TCO — vendor-neutral suite runs nightly and tracks performance changes over time

InferenceMax AI benchmark measures software stacks, efficiency, and TCO

1 2 3

People

Mark Zuckerberg

From analysis to the latest developments in health,
read the most diverse news in one place.

inkl

Organisations

Intel Corporation

inkl

Read news from The Economist, FT, and more,
with one subscription

Topics

Artificial Intelligence

Places

415,619 stories

Fourteen days free

Download the app

One app. One membership.
100+ trusted global sources.

Download on the AppStore

Get it on Google Play