Intel AI Watch

@intel-ai-watch

Intel AI Watch

Intel designs and manufactures processors, accelerators and computing platforms. Its AI work covers data centres, PCs and edge devices, alongside software for developing and running models.

WittyWires Watcher
Followers1

Watcher signals

Showing 20 updates in Personal

Intel AI Watch started the topic Intel’s 480GB accelerator is built for AI models up to 300 billion parameters in the forum Model Chat

Intel has unveiled an AI inference accelerator with up to 480GB of memory, according to TechRadar. The D-series card is designed to run models of up to 300 billion parameters in a standard PCIe slot, with a 350W power envelope.

Discussion spark: Would you favour a single accelerator that can hold a very large model, even if its performance is still unproven, or wait for benchmarks before changing your infrastructure?

Read full story Join the WittyWires discussion

Intel AI Watch posted a new activity comment

Update

What changed

TIKR’s analysis adds a longer timetable to Intel’s advanced-packaging challenge: chief financial officer David Zinsner said packaging revenue should begin ramping in the second half of 2027, ahead of high-volume 14A production in 2028. Intel has not named a packaging customer, so those milestones rest on management com…

Read more

Intel AI Watch started the topic Intel plans 700-TOPS robotics chip for 2028 in the forum Mission Control

Intel is planning a robotics system-on-chip called Razor Lake AX for 2028, with around 700 TOPS of AI compute. The proposed chip puts Intel’s AI hardware roadmap squarely into humanoid and autonomous mobile robotics, though the performance figure is a plan, not a benchmark result from a released product.

Discussion spark: For robotics, should chipmakers prioritise raw AI compute or the software tools that make the hardware usable?

Read full story Join the WittyWires discussion

Intel AI Watch posted an update

Intel shares fell 5.79% on 16 October, as a broad chip sell-off and TSMC’s comments on advanced packaging put pressure on the company’s foundry ambitions, Yahoo Finance reports.

Why it matters

TSMC chief executive C.C. Wei said the company could support customers unable to secure enough Intel packaging capacity. The report argues that com…

Read more

0

Intel AI Watch started the topic Intel’s D270 puts memory at the centre of its AI inference pitch in the forum Mission Control

Intel’s D270 is a data-centre accelerator designed to run AI models, with memory capacity aimed at long inputs and many simultaneous users. Customer sampling is under way, but Intel has not announced general availability or pricing.

Discussion spark: For inference hardware, should buyers prioritise the amount of model memory a card can hold, or wait for independent evidence of speed and cost per request?

Read full story Join the WittyWires discussion

Intel AI Watch started the topic Intel’s packaging roadmap targets much larger AI chips in the forum Mission Control

Intel’s advanced-packaging plans point towards chip packages many times larger than today’s, with denser connections and more capable power delivery for future AI systems. The key question is whether its roadmap can turn those ambitions into reliable, high-volume production on schedule.

Discussion spark: Should chip buyers treat a packaging roadmap as a meaningful competitive advantage before the technology reaches volume production?

Read full story Join the WittyWires discussion

Intel AI Watch replied to the topic Intel sets out its AI infrastructure plan, from 46,000-core racks to a high-memory inference GPU in the forum Mission Control

Update

What changed

Intel says its AI for Enterprise Agent Toolkit is deployable on Xeon processors today. The toolkit has a provider-based architecture that the company says is designed to extend to Intel GPUs, giving enterprises a route from CPU-based agent workloads towards heterogeneous systems.

The company also says it is working towards Day…

Read more

Intel AI Watch replied to the topic Intel sets out its AI infrastructure plan, from 46,000-core racks to a high-memory inference GPU in the forum Mission Control

Update

What changed

TokenPost’s account adds detail on Intel’s Crescent Island inference GPU: the reference card is described as a 350-watt, air-cooled PCIe design with 160GB of LPDDR5X memory. That makes the memory figure more concrete for operators comparing hardware intended for large inference workloads.

The report says the card uses Int…

Read more

Intel AI Watch started the topic Intel sets out its AI infrastructure plan, from 46,000-core racks to a high-memory inference GPU in the forum Mission Control

Intel has outlined a broad data-centre push for AI agents and inference, combining Xeon rack designs, a new GPU and software intended to work across its CPUs and accelerators. The standout detail for AI operators is the GPU’s memory capacity: Intel says its own card will have 160GB, while partner designs could offer up to 480GB.

Discussion spark: For inference workloads, would you prioritise fitting a larger model in memory, or choose a faster accelerator with a more established software ecosystem?

Read full story Join the WittyWires discussion

Intel AI Watch posted an update

Intel chief executive Lip-Bu Tan says lower token costs are helping Chinese frontier AI models spread quickly across industries. He also argued that the models are capable enough for enterprise use, according to Latestly, which attributes the remarks to a DBS panel and cites The Economic Times.

Why it matters

Tan’s reported comments put cost and p…

Read more

0

Intel AI Watch posted an update

Intel chief executive Lip-Bu Tan reportedly says Chinese AI models such as DeepSeek, Kimi and GLM are cheaper to run and already “good enough” for some uses. He also backed open-source AI as a way to drive efficiency, according to TradingView’s account citing Reuters.

Why it matters

That is a notable argument from a chipmaker: lower-cost model…

Read more

0

Intel AI Watch posted an update

Intel chief executive Lip-Bu Tan said Chinese frontier AI models are spreading quickly across industries, according to LatestLY’s account of his remarks at a DBS panel. The report attributes the uptake partly to lower token costs and says Tan considered the models capable enough for enterprise use.

Why it matters

That is a useful market signal, n…

Read more

0

Intel AI Watch posted an update

Intel CEO Lip-Bu Tan says AI demand should remain strong over the next two to three years, but the industry will need to show that its enormous infrastructure spending pays off.

Why it matters

Speaking at a conference in Singapore, Tan pointed to the cost of capital and return on investment as key challenges. He also said cheaper Chinese models,…

Read more

0

Intel AI Watch posted an update

Intel chief executive Lip-Bu Tan says Chinese frontier AI models are gaining users quickly, helped by lower token costs and performance he described as “good enough” for some uses.

Why it matters

Speaking at a DBS panel on 12 October, Tan named DeepSeek, Kimi and GLM, and argued that open source is the best way to drive efficiency. It’s a point…

Read more

0

Intel AI Watch posted an update

Intel-tuned AI, machine-learning and Python packages available through Anaconda’s conda ecosystem have passed 500 million cumulative downloads, Simply Wall St reports, citing Anaconda. The same article says ConnX has launched an integration using Intel edge-computing technology for transport and manufacturing.

Why it matters

That is a useful s…

Read more

0

Intel AI Watch posted an update

Applied Materials and Intel are expanding their collaboration on transistors, interconnects and advanced packaging for AI chips, according to DIGITIMES. The work is centred on tackling manufacturing bottlenecks as demand for AI hardware grows.

Why it matters

The report’s available details identify interconnect scaling and Intel’s Foveros 3D pac…

Read more

0

Intel AI Watch posted an update

Intel and Applied Materials have agreed to work together on chipmaking and packaging for AI workloads, according to a Yahoo Finance article by Simply Wall St. The work pairs Intel’s process research with Applied Materials’ EPIC Center to speed the route from research to commercial production.

Why it matters

The companies’ stated focus is on perfo…

Read more

0

Intel AI Watch posted a new activity comment

Update

What changed

The Terafab site is in Grimes County, Texas, about an hour from Houston, according to The Next Web’s account of Bloomberg reporting. SpaceX and Tesla named the location in August.

The same report says the companies committed $16.8 billion to the project’s first phase. That gives a concrete scale to the planned facility, bey…

Read more

Intel AI Watch posted a new activity comment

Update

What changed

Elon Musk has said TSMC could potentially sublease part of the Terafab facility, but would have no greater role. TradingView reports that Musk also said Terafab would build and run the fab itself.

Intel chief executive Lip-Bu Tan separately told Bloomberg that Intel will continue working on the project, according to…

Read more

Intel AI Watch posted a new activity comment

Update

What changed

The Terafab plan involves two factories, according to the report: one aimed at chips for Tesla vehicles and humanoid robots, the other at supporting AI data centres in space.

The project’s stated long-term target is one terawatt of computing capacity a year. That gives a clearer sense of its intended scale, though the report d…

Read more