AMD Acquires Taalas to Etch AI Models Into Silicon for Faster Inference
11 Sources
[1]
AMD acquires AI chip startup Taalas to boost inference performance by etching models into silicon
In AMD's latest bid to upset Nvidia's dominance in AI hardware, the House of Zen has acquired AI chip company Taalas, which bakes model weights directly into silicon in a process that promises to boost inference performance by an order of magnitude or more. The deal, announced at market close on
[2]
AMD deepens AI inference bet with Taalas deal as chip race heats up
Aug 6 (Reuters) - Advanced Micro Devices (AMD.O), opens new tab said on Thursday it would buy chip startup Taalas for an undisclosed amount, strengthening its technology as it taps into the rapidly growing market for AI inference workloads. Toronto-based Taalas develops specialized silicon
[3]
AMD buys chip startup that hardwires AI models into its silicon
Advanced Micro Devices is counting on its graphics processing units to drive the bulk of its data center growth as cloud companies snap up all the advanced AI chips they can find. But as the generative artificial intelligence boom approaches its fourth anniversary, it's becoming clear that GPUs
[4]
AMD buys Taalas to etch AI models into silicon
AMD is buying Taalas, a Toronto startup that etches a single AI model permanently into a chip. The reward is blistering speed. The catch is that the chip can then run only that one model. AMD announced the deal on Thursday, after the market closed, and did not say what it paid. Taalas, founded in
[5]
AMD acquires Taalas AI inference chip startup
The Toronto-based startup, founded in 2023, has raised $219 million and builds chips hardwired for specific AI models AMD $AMD announced Thursday it has entered into a definitive agreement to acquire Taalas, a Toronto-based startup that builds chips designed for AI inference. Financial terms were
[6]
AMD acquires Taalas to hardwire AI models into silicon
Advanced Micro Devices Inc. said today it has agreed to buy Taalas Inc., a Toronto startup that hardwires artificial intelligence models directly into silicon, in a deal that pushes the chipmaker deeper into the market for AI inference. Terms were not disclosed. AMD shares rose about 1.5%
[7]
AMD AI inference: AMD deepens AI inference bet with Taalas deal as chip race heats up
Toronto-based Taalas develops specialized silicon designed to reduce computing and memory bottlenecks in AI inference, the process of running trained AI models to generate responses or predictions. Advanced Micro Devices said on Thursday it would buy chip startup Taalas for an undisclosed
[8]
AMD Buys AI Chip Startup Taalas -- Here's Why It Matters To NVDA Investors - Advanced Micro Devices (NASDAQ:
Advanced Micro Devices (NASDAQ:AMD) said Thursday it will acquire Taalas, a Toronto startup that hardwires entire AI models directly into silicon, for an undisclosed amount. The deal targets inference, the business of running trained models, which AMD projects may grow more than 80% annually and
[9]
AMD Snaps Up Taalas Weeks After Cerebras Deal, Chasing Chips That Bake AI Models Into Silicon
Just a few weeks after AMD inked a deal with Cerebras to boost the inference capabilities of the upcoming Helios rack-scale system, it has acquired Taalas, a startup that designs chips specifically for inference-related workloads. Taalas bakes AI models directly into silicon chips, and AMD has
[10]
AMD Acquires Taalas to Boost AI Inference Performance and Next-Gen Compute Solutions
Taalas' technology will complement AMD's full-stack AI platform, including AMD Helios rackscale solutions, AMD Instinct GPUs, AMD EPYC CPUs, AMD ROCm software and the company's expanding AI ecosystem. AMD announced it has reached a definitive agreement to acquire Taalas, a pioneer in specialized
[11]
AMD to Buy AI Chip Startup Taalas
AMD has reached a deal to buy the AI chip startup Taalas at undisclosed terms. AMD said that the deal would strengthen its long-term AI ambitions by adding a differentiated inference technology and engineering expertise. Founded in 2023 and based in Toronto, Taalas optimizes inference dataflows,
Share
Copy Link
AMD acquired Toronto-based Taalas, an AI chip startup that hardwires model weights directly into silicon. The deal strengthens AMD's AI inference capabilities with technology that promises 48x faster performance than traditional GPUs, though chips remain locked to specific models.
AMD Acquires Taalas to Challenge Nvidia in AI Inference
AMD acquires Taalas, a Toronto-based startup that etches AI models into silicon, marking the chip giant's latest move to challenge Nvidia's dominance in AI hardware
1
2
. The deal, announced Thursday after market close, focuses on making high-performance AI inference services faster and cheaper to run. While AMD didn't disclose financial terms, the acquisition brings in a company that raised $219 million since its 2023 founding3
5
. The move mirrors Nvidia's $20 billion Groq licensing deal from December, signaling how seriously chipmakers now take specialized inference technology1
. Vamsi Boppana, AMD's senior vice president of AI, stated that "AMD is building a full-stack AI platform that gives customers the flexibility to deploy the right compute solutions for every AI workload"2
5
.
Source: Benzinga
How Model-Specific Integrated Circuits Work
Taalas developed what it calls model-specific integrated circuits that hardwire AI models into silicon rather than loading weights from memory
1
4
. The chips eliminate compute and memory bottlenecks by merging memory and compute functions directly into the silicon2
4
. According to Taalas co-founder Ljubisa Bajic, this approach eliminates the need for high-bandwidth memory, 3D stacking, or liquid cooling4
. The processors contain two main regions: a mask-ROM recall fabric where model weights are etched, and an SRAM recall fabric for KV caches and fine-tuning adapters1
. Taalas manufactured its first test chip, the HC1, on TSMC's 6nm process technology in February1
.
Source: The Register
Breakthrough Inference Performance Numbers
The HC1 chip delivered remarkable inference performance when running Meta Llama 3.1 8B, achieving 16,960 tokens per second—48x faster than Nvidia GPUs and 8.5x faster than Cerebras accelerators at the time of announcement
1
3
. Taalas claims its technology can produce output for specific models thousands of times faster than traditional GPUs, though this comes at the cost of flexibility3
. The company's second-generation HC2 chip, due this summer, aims to support 20 billion parameters per chip1
. For larger models, weights can be distributed across multiple accelerators using pipeline parallelism—meaning just 50 accelerators would support a trillion-parameter model1
.The Trade-Off: Speed Versus Flexibility
The technology's primary limitation is that once deployed, chips remain locked to a single model
3
4
. Any significant change beyond LoRA adapters requires a silicon re-spin, which adds cost and time1
. However, Taalas mitigates this constraint by requiring changes to only two metal layers rather than starting from scratch, allowing new models to be etched in about two months instead of six1
4
. Taalas notes that fewer than a handful of a chip's hundred-plus layers need alteration from one design to another5
. The company claims etching model weights into silicon costs 100x less than training a frontier model1
. This positions the technology for AI model developers, infrastructure providers, and inference services rather than general enterprise customers1
.
Source: Wccftech
Related Stories
Integration Into AMD's AI Hardware Portfolio
AMD plans to integrate Taalas technology into its AI accelerator roadmap and develop system-level solutions using AMD Instinct GPUs, EPYC processors, and Helios rack-scale platform
2
5
. The company appears positioned to pair Instinct-based Helios racks with Taalas chips in a disaggregated architecture where compute-heavy prompt processing runs on GPUs while token generation offloads to Taalas accelerators1
. AMD could adopt a deployment strategy where customers initially validate models on Instinct accelerators before transitioning to Taalas accelerators for production1
. This approach offers far better space and power efficiency than Nvidia's LPX systems, which would need dozens of GPUs and at least 2,000 Groq LPUs to serve the same trillion-parameter model1
.Strategic Context in AMD's Acquisition Spree
The Taalas deal continues AMD's aggressive expansion of its AI capabilities through acquisitions
2
3
. AMD bought MK1, an AI software startup specializing in high-speed inference, in November2
. The company purchased MEXT in June and added FastFlowLM to its Artificial Intelligence Group in July2
. In 2024, AMD paid $665 million for Silo AI and acquired ZT Systems for $4.9 billion, which provided the technical foundation for its rack-scale products3
5
. AMD reported record revenue of $11.5 billion in its most recent quarter, with its data center segment more than doubling year-over-year to $6.7 billion5
. The company has secured Helios commitments from major customers including Meta, Microsoft, OpenAI, and Anthropic1
5
. AMD stock rose approximately 1.5% following the announcement5
.References
Summarized by
Navi
[1]
[4]
Related Stories
AMD's Strategic AI Acquisitions: Brium, Untether AI Team, and Enosemi Bolster Competitive Edge Against Nvidia
05 Jun 2025•Technology
AMD Acquires Neuralink Veterans' AI Startup MK1 to Enhance Instinct GPU Performance
11 Nov 2025•Business and Economy
Taalas raises $169 million to build custom AI chips that hardwire models into silicon
19 Feb 2026•Technology
Recent Highlights
1
OpenAI Solves Navier-Stokes Problem With AI, Sparking Existential Crisis in Mathematics
Science and Research
2
OpenAI Reveals AI Models Hiding Mistakes and Taking Unauthorized Actions in 6 New Misalignment Cases
Technology
3
OpenAI AI Agents Attack RubyGems With 2,000+ Malicious Packages, Attempt API Key Theft
Technology
Recent Highlights
Today's Top Stories
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.
