News
IndustryTSMC’s grip on advanced chipmaking will face competitionIndustryCadence joins European Chip Design Platform to support semiconductor startups and SMEsMaking chipsMarc Hijink on ASML’s Culture, Supply Chain, and GeopoliticsProcessorsIntel's post-Nova Lake CPUs take shape as Razor Lake desktop and laptop variants surfaceAI chipsHow Google’s New AI Chip Software Helps It Battle NvidiaIndiaMore IC Packaging Plants Emerge In IndiaIndustryUS Tightens China AI Chip Access as Nvidia, TSMC and Semiconductor Stocks SlideDesignAxelera AI moves into HPC with TitaniaMaking chipsChinese envoy says chip equipment export curbs threaten EU-China trade thawIndiaSEMICON India 2026 Day 2 Sees 25 MoUs, Strategic Collaborations Across Semiconductor Value ChainIndustryTSMC’s grip on advanced chipmaking will face competitionIndustryCadence joins European Chip Design Platform to support semiconductor startups and SMEsMaking chipsMarc Hijink on ASML’s Culture, Supply Chain, and GeopoliticsProcessorsIntel's post-Nova Lake CPUs take shape as Razor Lake desktop and laptop variants surfaceAI chipsHow Google’s New AI Chip Software Helps It Battle NvidiaIndiaMore IC Packaging Plants Emerge In IndiaIndustryUS Tightens China AI Chip Access as Nvidia, TSMC and Semiconductor Stocks SlideDesignAxelera AI moves into HPC with TitaniaMaking chipsChinese envoy says chip equipment export curbs threaten EU-China trade thawIndiaSEMICON India 2026 Day 2 Sees 25 MoUs, Strategic Collaborations Across Semiconductor Value Chain
EveryChip
Colossus MK2 GC200

Colossus MK2 GC200

7nm AI processor chip by Graphcore

The GC200 is a second-generation artificial intelligence processor designed by the United Kingdom firm Graphcore. Built on a seven-nanometre process, it powers the M2000 system and supports large-scale parallel computing for machine learning tasks.

Share on WhatsApp

  • Contains 59.4 billion transistors per chip
  • Built on a 7nm manufacturing process
  • Powers the M2000 IPU Machine system
  • Supports scaling up to 16 exaflops
  • Designed specifically for AI workloads

What it is

The Colossus MK2 GC200 is a specialised silicon processor created for artificial intelligence workloads. Graphcore, a startup based in the United Kingdom, designed this chip as part of its second generation of hardware. It serves as the core component of the M2000 IPU Machine. The company states that this architecture offers a distinct alternative to traditional graphics processing units used in AI. Each individual chip contains 59.4 billion transistors and is manufactured using a seven-nanometre process node. This technology aims to handle complex calculations required for modern machine learning models efficiently.

How it came about

Graphcore launched the GC200 in November 2026, marking its first new hardware release in nearly two years. The development comes as the demand for advanced AI processing grows across sectors like healthcare and financial services. CEO Nigel Toon highlighted that the company remains focused solely on AI silicon rather than diversifying into general-purpose CPUs or GPUs. This strategic choice reflects their belief that dedicated processors provide better efficiency for specific computational tasks. The launch coincided with significant competition from larger industry players developing their own AI accelerators.

Inside the chip

The chip employs an Intelligence Processing Unit (IPU) architecture, which Graphcore argues is more efficient than standard GPU designs for AI tasks. It enables massive parallelism, allowing thousands of these processors to connect together. The design supports systems with up to 16 exaflops of computing power and petabytes of memory. This scalability helps manage models with trillions of parameters. The company claims this approach reduces the total cost of ownership significantly compared to other solutions. The internal structure is optimised for the specific mathematical operations common in deep learning algorithms.

How it was made

Manufactured on a seven-nanometre process, the GC200 integrates 59.4 billion transistors onto a single die. While the specific foundry is not named in the provided text, the advanced node indicates high-density production techniques. The chip is not sold as a standalone component. Instead, it is packaged within the M2000 IPU Machine, which houses four of these processors. This integration allows for immediate deployment into existing data centre infrastructure without requiring custom board design by the end user.

Where it went

The GC200 powers the M2000 system, which targets industries such as financial services, healthcare, and technology. It is designed for applications where artificial intelligence is used to solve complex challenges, including medical breakthroughs and cybersecurity. Early access customers have begun receiving the hardware, with wider availability planned for the end of the year. The system supports tasks ranging from navigation improvements for autonomous vehicles to large-scale data analysis. Its ability to scale makes it suitable for both small clusters and vast parallel processing networks.

Versions and family

The GC200 is the second generation of Graphcore’s hardware line. It succeeds the previous generation of IPU chips, though specific model names for the predecessor are not detailed here. The chip is exclusively available as part of the M2000 IPU Machine. There are no plans to sell the GC200 separately. The M2000 itself can be scaled by connecting multiple units, creating larger systems that maintain the same architectural principles. This modular approach allows users to expand their computing resources as needed.

Why it matters

Graphcore positions the GC200 as a competitor to major players like Nvidia and Intel in the AI hardware market. The company raised substantial funding to support this effort, valuing the business at nearly two billion dollars. By focusing entirely on AI silicon, Graphcore aims to capture a share of the growing market for specialised processors. The claim of lower ownership costs could influence how enterprises build their compute infrastructure. If successful, this approach might shift industry standards away from general-purpose GPUs for certain AI workloads.

Problems and reception

The chip has attracted early access customers, including major technology firms like Microsoft. Industry observers note the intense competition in the AI chip sector, with rivals releasing powerful alternatives simultaneously. Graphcore’s leadership expresses confidence that their focused strategy will yield profitability despite the presence of larger competitors. The reception hinges on whether the promised efficiency gains and cost reductions materialise in real-world deployments. The company continues to invest heavily in technology and personnel to support this rollout.

Questions people ask about Colossus MK2 GC200

Can I buy the GC200 chip on its own?

No, Graphcore does not plan to sell the GC200 separately. It is only available as part of the M2000 IPU Machine, which contains four of these chips. This integrated approach simplifies deployment for customers.

What is the transistor count of the GC200?

Each GC200 chip features 59.4 billion transistors. This high density is achieved using a seven-nanometre manufacturing process, allowing for significant computational power in a compact form factor.

How does the IPU differ from a GPU?

Graphcore states that its IPU architecture is more efficient for AI tasks than traditional GPUs. It is designed specifically for machine learning workloads, offering lower cost of ownership and better scalability for large models.

Who are the main competitors for this chip?

The GC200 competes against AI processors from major companies like Nvidia and Intel. These firms also produce hardware for artificial intelligence, creating a crowded market for specialised silicon solutions.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Colossus MK2 GC200 yet, so the facts on this page come from the sources listed above: datacenterdynamics.com/en/news/graphcore-announces-second-gen-ai-chip-, au.finance.yahoo.com/news/graphcore-unveils-g200-chip-expandable-07322. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryChip wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the chip of the day, new pages on EveryChip, and the conferences and anniversaries of the week ahead.