Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesSoftBank Group Corp. acquired Graphcore in July 2024, making the British AI-processor designer a wholly owned subsidiary while keeping its name and Bristol base. Graphcore did not disclose the price; contemporary reports put it at roughly $400 million to $500 million, far below the company’s reported late-2020 valuation of about $2.8 billion. The deal was both a rescue of a cash-constrained chip business and a strategic purchase of AI-processor technology, software and engineering talent.
What happened in the Graphcore acquisition?
Graphcore dated its official announcement July 11, 2024. The buyer was SoftBank Group Corp., not the separate telecommunications operating company SoftBank Corp. Graphcore became wholly owned by SoftBank Group, retained its name and continued operating from Bristol. Nigel Toon remained chief executive at the time.
Graphcore’s announcement did not state a consideration figure. Contemporary reports cited different estimates—about $400 million in one account and about $500 million in another—so neither should be treated as the definitive purchase price. Graphcore’s official account is available in its acquisition announcement.
This was not a shutdown, a rename or an announced merger into Arm. It was a change of ownership intended to give Graphcore a stronger financial and strategic base for developing AI compute.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →#1 Best Overall
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
Why Graphcore needed a buyer
Strong technology, difficult business conditions
Graphcore designed specialized processors for artificial-intelligence workloads and attracted major investment, including funding associated with Microsoft and Sequoia Capital. Nigel Toon later described more than $600 million in equity funding; contemporary coverage also referred to roughly $700 million of investment. The company reached a reported private valuation of approximately $2.8 billion in late 2020.
That valuation reflected the AI investment boom, not a guarantee of durable revenue. Graphcore won customers and deployments, but it struggled to build the scale, financing runway and software ecosystem needed to make its architecture a widely adopted alternative to Nvidia’s accelerator platform.
The retrenchment before the sale
Contemporary reporting said Graphcore cut about 20% of its workforce, leaving roughly 500 employees, and reduced its geographic footprint. Operations were reportedly closed or scaled back in countries including Norway, Japan and South Korea. These measures indicate pressure on costs and capital, but they do not mean the underlying technology had no value.
An AI-chip company must solve several problems at once:
- Silicon must deliver useful performance on real customer workloads.
- Compilers, libraries and framework support must let developers use the hardware without rewriting entire systems.
- Customers need dependable supply, system integration and long-term support.
- The company must finance successive chip generations while competing with a much larger incumbent.
Nvidia’s advantage therefore extended well beyond chip specifications. CUDA, optimized libraries, cloud availability, server vendors, developer familiarity and a large installed base created switching costs that Graphcore had to overcome.
Rank #2
- High-Performance AI Processing: The MX3 is designed to handle the most demanding AI computer vision workloads, delivering exceptional performance and efficiency.
- Flexible Integration: The MX3 can be easily integrated into your existing systems via its M.2 M-key form factor and support for Linux operating systems.
- Energy Efficient: The MX3 is designed to provide high performance while minimizing power consumption.
- Comprehensive Software Development Kit (SDK): The MX3 is supported by a comprehensive SDK that simplifies development and deployment.
- Hardware compatability: The MX3 is compatible with the PCI-SIG M.2 M-key 2280 Specification. It can be used with the Raspberry Pi 5 with a M-key 2280 HAT.
What Graphcore built: the IPU and Poplar software
IPU architecture
Graphcore’s central product was the Intelligence Processing Unit (IPU), a processor designed specifically for highly parallel machine-learning workloads. Its architecture emphasized many independent cores, substantial on-chip SRAM, high internal memory bandwidth and a graph-oriented execution model. Specialized links connected processors so larger systems could be built from multiple IPUs.
Graphcore’s software stack, known as Poplar, was essential to that design. Poplar maps computational graphs and data movement onto the IPU. In practice, the compiler, framework integrations, model support and debugging tools matter as much as the silicon: a theoretically powerful processor is of limited commercial use if developers cannot port and operate their models efficiently.
How an IPU differs from a GPU
A GPU is a broadly programmable accelerator with an exceptionally mature AI ecosystem. An IPU is more specialized around Graphcore’s parallel, graph-based execution model. The trade-off is straightforward:
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minute| Dimension | Graphcore IPU approach | Nvidia GPU approach |
|---|---|---|
| Execution model | Graph-oriented execution across many independent cores | General-purpose parallel accelerator model with extensive AI libraries |
| Memory emphasis | Large on-chip SRAM and very high local bandwidth | Broad support for external high-bandwidth memory and established system designs |
| Software position | Poplar compiler and Graphcore-specific integrations | CUDA, libraries, frameworks and a large developer ecosystem |
| Commercial strength | Differentiated architecture requiring workload fit and software adoption | Wide availability, system-vendor support and extensive customer familiarity |
Neither architecture wins every workload automatically. Memory movement, precision, sparsity, model partitioning, compiler quality, interconnect, batch size and utilization can change the result. Peak TFLOPS alone cannot establish faster training, lower inference cost or better total cost of ownership.
Colossus MK2 specifications
Contemporary technical coverage reported the following figures for Graphcore’s Colossus MK2 family. They are hardware specifications, not proof of application-level superiority over Nvidia.
Rank #3
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
| Specification | Reported figure |
|---|---|
| Transistors | Approximately 59.4 billion |
| Independent cores | 1,472 |
| Simultaneous multithreading | Up to 8,832 parallel threads |
| On-chip SRAM | 900 MB |
| Aggregate on-chip bandwidth | Approximately 47.5 TB/s |
| IPU links | Ten links for scaling between processors |
| MK2 C600 FP8 | 560 TFLOPS |
| MK2 C600 FP16 | 280 TFLOPS |
| MK2 C600 FP32 | 70 TFLOPS |
| MK2 C600 power | Approximately 185 W |
See Graphcore’s MK2 C600 product reference and the contemporaneous technical and business report for the cited figures.
Why the sale price was far below the old valuation
A private valuation and an acquisition price answer different questions. The approximately $2.8 billion figure was a reported financing-era valuation in late 2020, when investors placed a premium on independent AI-chip challengers. The later transaction was negotiated after Graphcore had faced funding pressure, workforce reductions, fierce competition and the practical difficulty of turning deployments into a scaled platform business.
Free tools Windows power users keep installed
One-click scans. No signup required.
A strategic buyer may value selected assets—processor designs, software, patents, engineering teams, customer knowledge and future options—without paying the earlier price for the entire independent-company story. Funding needs, customer concentration, market timing and the loss of the “next Nvidia” narrative can all compress a sale price.
Accordingly, the reported $400 million and $500 million figures should be read as estimates from contemporary coverage, not as a disclosed final consideration. Graphcore’s announcement itself did not publish the amount.
What SoftBank bought beyond a chip
The transaction gave SoftBank control of several strategically relevant assets:
Rank #4
- 48GB AI graphics accelerator
- Graphcore’s IPU architecture and related semiconductor designs.
- The Poplar compiler and graph-execution software stack.
- Chip design, verification, compiler and systems engineering expertise.
- AI-system and data-center deployment knowledge.
- Customer relationships and experience operating real workloads.
- A UK-based organization that could be funded for a longer development cycle than venture financing allowed.
Graphcore described the deal as a platform for building the “next generation of AI compute.” A SoftBank representative connected next-generation semiconductors and compute systems with the group’s AGI ambitions. Those statements establish strategic intent, but they do not disclose a detailed product roadmap.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Why SoftBank might want that position
- AI-compute exposure: owning an accelerator designer gives SoftBank a direct position in the hardware layer of the AI economy.
- Engineering talent: Graphcore combines processor architecture, compilers and AI-systems skills that are difficult to assemble quickly.
- Differentiated intellectual property: SoftBank can develop an alternative architecture rather than relying exclusively on Nvidia.
- Potential Arm adjacency: Graphcore could potentially work with Arm, but no formal Arm integration was announced.
- A longer investment horizon: a strategic parent can finance research and product development beyond a venture-backed company’s immediate fundraising cycle.
- Infrastructure optionality: Graphcore could become one component of broader SoftBank data-center and AI initiatives.
These are strategic possibilities, not proof that SoftBank had committed to a specific combined product or commercial plan.
What has happened under SoftBank?
First-party updates through August 16, 2026 show continued operation and expansion rather than an IP-only wind-down.
- Graphcore said it was approaching 1,000 employees.
- It opened development centers in Austin, Texas, and Bengaluru, India.
- It expanded activity in Taiwan, Poland, Cambridge and London.
- It planned to move into a purpose-built Bristol headquarters in September 2026.
- It announced a Taipei office and engineering lab on August 3, 2026, citing continued investment in Taiwan and semiconductor supply-chain relationships.
- Co-founder and executive chair Nigel Toon stepped down effective July 31, 2026; Marcus McElroy took leadership.
The workforce and site announcements are consistent with SoftBank preserving and investing in Graphcore as an operating company. They do not establish revenue growth, profitability, production volume or market share. The leadership change should likewise be reported as a fact, not automatically classified as either a crisis or a success. Details appear in Graphcore’s July 31, 2026 announcement and Taipei announcement.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why Graphcore had not displaced Nvidia
Software and developer adoption
Portability is a central hurdle. Customers often build around CUDA libraries, established frameworks and familiar operational tools. Graphcore’s Poplar stack had to provide compelling model support, compiler performance and debugging workflows while convincing developers to learn a different platform.
Best Value
- Professional AI & Creator Workstation: AMD Radeon AI PRO R9700 GPU with 32GB GDDR6 is engineered for AI development, professional content creation, and compute-intensive workloads.
- Massive 32GB Memory Capacity: 32GB of GDDR6 memory on a 256-bit bus provides ample bandwidth for large AI models, 8K video editing, and complex 3D rendering.
- Advanced RDNA 4 with AI Accelerators: 64 Compute Units with 3rd Gen Ray Tracing and dedicated 2nd Gen AI Accelerators for groundbreaking AI performance and visual computing.
- Professional Blower Cooling: Efficient single blower design exhausts heat directly out of the chassis, ideal for multi-GPU workstation and server configurations.
- Enterprise-Grade Thermal Solution: Vapor chamber heatsink with industrial Honeywell PTM7950 thermal interface material ensures reliable cooling under sustained professional loads.
Availability and confidence
AI buyers also assess supply, system vendors, cloud access, support contracts and a supplier’s ability to deliver several product generations. A technically attractive accelerator can lose deals if customers fear shortages, uncertain maintenance or an abrupt end to the roadmap.
Specifications versus production results
The MK2 figures describe arithmetic and memory capabilities under specified conditions. They do not by themselves measure tokens per second, training time, latency, utilization, energy per inference or total system cost on a customer’s model. Those outcomes require workload-specific, end-to-end comparisons.
How to judge whether the acquisition worked
Investors and infrastructure buyers should look for evidence that connects SoftBank’s investment to commercial outcomes:
- New Graphcore chip generations and products that reach production.
- Named customer deployments at meaningful scale.
- Reliable manufacturing capacity and system availability.
- Revenue, order volume or other evidence of repeatable demand.
- Support for current AI frameworks and widely used models.
- Growing developer adoption of Poplar and associated tools.
- Benchmarks on representative production workloads, not only peak arithmetic figures.
- Cloud availability, data-center partnerships or integration with major server platforms.
- Evidence that hiring growth produces commercially useful output.
- A clear role for Graphcore—independent product company, partner to Arm and other SoftBank businesses, or internal technology group.
Assessment
SoftBank bought Graphcore at a distressed valuation because the company’s technology and people still had strategic value even though Graphcore had not created a commercially durable Nvidia alternative. The acquisition supplied capital, time and ownership of an AI-processor platform that SoftBank could develop alongside its broader semiconductor and infrastructure ambitions.
As of August 2026, Graphcore’s expansion and continued separate identity show preservation and investment, not proven market success. The most accurate description is a second chance and a strategic AI-compute bet—not evidence that Graphcore defeated Nvidia or that its IPU became a mainstream GPU replacement.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
