Graphcore
British AI-accelerator designer that built the IPU chip and open Poplar software stack, now a wholly owned SoftBank subsidiary developing next-generation compute for SoftBank's AI infrastructure push.
Total funding—
Valuation—
Employees1,000
Founded2016
HeadquartersBristol, United Kingdom
Bristol, United KingdomStatusPrivate
Product
Selected offerings — the AI products recorded here, not the vendor's full catalogue.
Inside Bow Pod systems
-
Massively parallel processor purpose-built for machine intelligence, paired with the Poplar SDK; the current Colossus MK2 GC200 IPU is packaged into the IPU-M2000 system delivering 1 petaFLOP of AI compute.
Closest alternatives 1
Similar products 36
H
HuaweiPrivateIPU (Intelligence Processing Unit) and Atlas do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; scale not recorded for either.4 of 4 capabilitiesShared product typeCompareA
AMDPublic · $785.1BIPU (Intelligence Processing Unit) and AMD Instinct do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.4 of 4 capabilitiesShared product typeCompareG
GroqPrivate · $3.5BIPU (Intelligence Processing Unit) and Groq LPU (GroqChip) do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareCS
Cerebras SystemsPublic · $44.4BIPU (Intelligence Processing Unit) and Cerebras WSE-3 Turbo (WSE-3T) do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareAA
Axelera AIPrivate · $450M raisedIPU (Intelligence Processing Unit) and Axelera Metis AIPU do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareQ
QualcommPublic · $174.9BIPU (Intelligence Processing Unit) and Qualcomm Cloud AI 100 Ultra do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareS
SambaNovaPrivate · $11BIPU (Intelligence Processing Unit) and SambaNova RDU do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareT
TenstorrentPrivate · $2.6BIPU (Intelligence Processing Unit) and Blackhole do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareI
IntelPublic · $463.7BIPU (Intelligence Processing Unit) and Intel Gaudi AI Accelerator do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareHR
Horizon RoboticsPublic · $8.8BIPU (Intelligence Processing Unit) and Journey 6 Series do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareDT
Dell TechnologiesPublic · $360.7BIPU (Intelligence Processing Unit) and Dell Pro Max AI Desktop do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareSH
SK HynixPublic · $874.5BIPU (Intelligence Processing Unit) and HBM3E do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareB
BroadcomPublic · $1.7TIPU (Intelligence Processing Unit) and Broadcom custom silicon do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.1 of 4 capabilitiesShared product typeCompareC
CiscoPublic · $433.3BIPU (Intelligence Processing Unit) and Cisco AI PODs do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareMT
Marvell TechnologyPublic · $214.9BIPU (Intelligence Processing Unit) and Marvell custom AI silicon do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.1 of 4 capabilitiesShared product typeCompareAH
Arm HoldingsPublic · $282.8BIPU (Intelligence Processing Unit) and Arm CSS for Mobile 2 do comparable work on accelerator silicon; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareAM
Applied MaterialsPublic · $380.7BIPU (Intelligence Processing Unit) and SEMVision H20 do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.1 of 4 capabilitiesShared product typeCompareHT
Hailo TechnologiesAcquired ($340M deal)IPU (Intelligence Processing Unit) and Hailo AI accelerators do comparable work on accelerator silicon; Graphcore's scale not recorded; Hailo Technologies is acquired, with no independent scale.2 of 4 capabilitiesShared product typeCompareM
ModularAcquired ($3.1B deal)IPU (Intelligence Processing Unit) and Mojo do comparable work on GPU programming; both also serve buyers who need to program GPU accelerators; Graphcore's scale not recorded; Modular is acquired, with no independent scale; ships developer tool and infrastructure service rather than the same layer.2 of 4 capabilitiesDifferent layerCompareD
DecartPrivate · $4BIPU (Intelligence Processing Unit) and Decart Optimization Stack do comparable work on GPU programming and model training; both also serve buyers who need to program GPU accelerators; Graphcore's scale not recorded; ships API service, application and 2 more rather than the same layer.3 of 4 capabilitiesDifferent layerCompareA
AmazonPublic · $2.8TIPU (Intelligence Processing Unit) and AWS Trainium do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareA
Alphabet (Google)Public · $4.2TIPU (Intelligence Processing Unit) and Google Cloud TPU do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareA
ASMLPublic · $670.1BIPU (Intelligence Processing Unit) and ASML DUV lithography systems both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareK
KioxiaPublic · $177.9BIPU (Intelligence Processing Unit) and KIOXIA CM9-V Series both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareT
TSMCPublic · $2TIPU (Intelligence Processing Unit) and TSMC 3DFabric both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareC
CrusoePrivate · $10BIPU (Intelligence Processing Unit) and Crusoe Spark both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareAN
Arista NetworksPublic · $246.4BIPU (Intelligence Processing Unit) and 7060X5 Series both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareSE
Samsung ElectronicsPublic · $1.2TIPU (Intelligence Processing Unit) and Samsung Datacenter SSD both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareL
LambdaPrivate · $1.5B raisedIPU (Intelligence Processing Unit) and Lambda Cloud do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships infrastructure service and model API rather than the same layer.2 of 4 capabilitiesDifferent layerCompareNG
Nebius GroupPublic · $61.6BIPU (Intelligence Processing Unit) and Nebius AI Cloud do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships developer tool, infrastructure service and 1 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompareR
RunPodPrivateIPU (Intelligence Processing Unit) and Pods do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; scale not recorded for either; ships developer tool, infrastructure service and 2 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompare4
4ParadigmPublic · $2BIPU (Intelligence Processing Unit) and 4Paradigm SageOne both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareC
CoreWeavePublic · $48.5BIPU (Intelligence Processing Unit) and CoreWeave AI Cloud do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships AI agent, infrastructure service and 1 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompareH
HumainPrivateIPU (Intelligence Processing Unit) and HUMAIN COMPUTE do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; scale not recorded for either; ships API service, agent platform and 1 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompareS
SenseTimePublic · $7.7BIPU (Intelligence Processing Unit) and SenseCore do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareA
AlibabaPublic · $273.4BIPU (Intelligence Processing Unit) and Alibaba Cloud Elastic GPU Service do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships AI agent, API service and 4 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompare
HuaweiPrivateIPU (Intelligence Processing Unit) and Atlas do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; scale not recorded for either.4 of 4 capabilitiesShared product typeCompareAAMDPublic · $785.1BIPU (Intelligence Processing Unit) and AMD Instinct do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.4 of 4 capabilitiesShared product typeCompareGGroqPrivate · $3.5BIPU (Intelligence Processing Unit) and Groq LPU (GroqChip) do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareCSCerebras SystemsPublic · $44.4BIPU (Intelligence Processing Unit) and Cerebras WSE-3 Turbo (WSE-3T) do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareAAAxelera AIPrivate · $450M raisedIPU (Intelligence Processing Unit) and Axelera Metis AIPU do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareQQualcommPublic · $174.9BIPU (Intelligence Processing Unit) and Qualcomm Cloud AI 100 Ultra do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareSSambaNovaPrivate · $11BIPU (Intelligence Processing Unit) and SambaNova RDU do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareTTenstorrentPrivate · $2.6BIPU (Intelligence Processing Unit) and Blackhole do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareIIntelPublic · $463.7BIPU (Intelligence Processing Unit) and Intel Gaudi AI Accelerator do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareHRHorizon RoboticsPublic · $8.8BIPU (Intelligence Processing Unit) and Journey 6 Series do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareDTDell TechnologiesPublic · $360.7BIPU (Intelligence Processing Unit) and Dell Pro Max AI Desktop do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareSHSK HynixPublic · $874.5BIPU (Intelligence Processing Unit) and HBM3E do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareBBroadcomPublic · $1.7TIPU (Intelligence Processing Unit) and Broadcom custom silicon do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.1 of 4 capabilitiesShared product typeCompareCCiscoPublic · $433.3BIPU (Intelligence Processing Unit) and Cisco AI PODs do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareMTMarvell TechnologyPublic · $214.9BIPU (Intelligence Processing Unit) and Marvell custom AI silicon do comparable work on accelerator silicon; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.1 of 4 capabilitiesShared product typeCompareAHArm HoldingsPublic · $282.8BIPU (Intelligence Processing Unit) and Arm CSS for Mobile 2 do comparable work on accelerator silicon; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareAMApplied MaterialsPublic · $380.7BIPU (Intelligence Processing Unit) and SEMVision H20 do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.1 of 4 capabilitiesShared product typeCompareHTHailo TechnologiesAcquired ($340M deal)IPU (Intelligence Processing Unit) and Hailo AI accelerators do comparable work on accelerator silicon; Graphcore's scale not recorded; Hailo Technologies is acquired, with no independent scale.2 of 4 capabilitiesShared product typeCompareMModularAcquired ($3.1B deal)IPU (Intelligence Processing Unit) and Mojo do comparable work on GPU programming; both also serve buyers who need to program GPU accelerators; Graphcore's scale not recorded; Modular is acquired, with no independent scale; ships developer tool and infrastructure service rather than the same layer.2 of 4 capabilitiesDifferent layerCompareDDecartPrivate · $4BIPU (Intelligence Processing Unit) and Decart Optimization Stack do comparable work on GPU programming and model training; both also serve buyers who need to program GPU accelerators; Graphcore's scale not recorded; ships API service, application and 2 more rather than the same layer.3 of 4 capabilitiesDifferent layerCompareAAmazonPublic · $2.8TIPU (Intelligence Processing Unit) and AWS Trainium do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareAAlphabet (Google)Public · $4.2TIPU (Intelligence Processing Unit) and Google Cloud TPU do comparable work on accelerator silicon and model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.3 of 4 capabilitiesShared product typeCompareAASMLPublic · $670.1BIPU (Intelligence Processing Unit) and ASML DUV lithography systems both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareKKioxiaPublic · $177.9BIPU (Intelligence Processing Unit) and KIOXIA CM9-V Series both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareTTSMCPublic · $2TIPU (Intelligence Processing Unit) and TSMC 3DFabric both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareCCrusoePrivate · $10BIPU (Intelligence Processing Unit) and Crusoe Spark both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareANArista NetworksPublic · $246.4BIPU (Intelligence Processing Unit) and 7060X5 Series both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareSESamsung ElectronicsPublic · $1.2TIPU (Intelligence Processing Unit) and Samsung Datacenter SSD both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.Shared product typeCompareLLambdaPrivate · $1.5B raisedIPU (Intelligence Processing Unit) and Lambda Cloud do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships infrastructure service and model API rather than the same layer.2 of 4 capabilitiesDifferent layerCompareNGNebius GroupPublic · $61.6BIPU (Intelligence Processing Unit) and Nebius AI Cloud do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships developer tool, infrastructure service and 1 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompareRRunPodPrivateIPU (Intelligence Processing Unit) and Pods do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; scale not recorded for either; ships developer tool, infrastructure service and 2 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompare44ParadigmPublic · $2BIPU (Intelligence Processing Unit) and 4Paradigm SageOne both serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareCCoreWeavePublic · $48.5BIPU (Intelligence Processing Unit) and CoreWeave AI Cloud do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships AI agent, infrastructure service and 1 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompareHHumainPrivateIPU (Intelligence Processing Unit) and HUMAIN COMPUTE do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; scale not recorded for either; ships API service, agent platform and 1 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompareSSenseTimePublic · $7.7BIPU (Intelligence Processing Unit) and SenseCore do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded.2 of 4 capabilitiesShared product typeCompareAAlibabaPublic · $273.4BIPU (Intelligence Processing Unit) and Alibaba Cloud Elastic GPU Service do comparable work on model training; both also serve buyers who need to buy data-centre AI compute; Graphcore's scale not recorded; ships AI agent, API service and 4 more rather than the same layer.2 of 4 capabilitiesDifferent layerCompareClosest alternatives are pairs a vendor page itself compares. Similar products are matched from what each product does; nothing in that group is asserted as a rivalry.
Sources
- Employees
-
Third-party database · 29 Aug 2026
Aggregator estimate, ~1,000 as of 2026; rounded. No post-acquisition valuation or funding figure is public -- Graphcore no longer raises independently -- so both are left null rather than reused from its pre-acquisition 2020 round.
- Founded
-
Founded 2016 in Bristol.
- Status
-
SoftBank Group completed its acquisition of Graphcore, 11 July 2024, making it a wholly owned subsidiary. SoftBank is not itself a record in this directory -- one acquisition does not admit an acquirer on its own (see /methodology/) -- so no acquired edge is minted and this stays a status fact in prose plus this provenance line, the same treatment as Haptik and Unbabel. Graphcore ships under its own name today (the Izanagi chip programme for SoftBank's Stargate infrastructure), which is why it stays status: active per the ghost-company test rather than being retired.
Open Graphcore in the directory
Something wrong here? Send a correction — quote this company id: graphcore.