Coopitition is a Pretty Old Story in Technology

“Coopitition” happens frequently in many markets, as competitors find they also cooperate with their rivals. But value chain participants also often move into other parts of the chain, meaning customers become competitors. 

Original Supplier	Customer (Later Competitor)	What the Customer Originally Bought	How/When Customer Became Competitor	Nature of Competition
Intel	Apple (M1/M2/M3 chips)	x86 CPUs for Mac computers	2020: Apple launched Apple Silicon and began replacing Intel CPUs in all Macs	Direct chip-level competitor; vertically integrated into SoC design
Qualcomm	Samsung, Huawei (HiSilicon), Apple (modems underway)	Mobile baseband chips	Samsung & Huawei developed in-house modems; Apple pursuing own modems	Reduced reliance on Qualcomm; internal components compete directly
NVIDIA	Amazon AWS, Google, Microsoft Azure	GPUs for cloud AI workloads	Clouds built custom AI chips (AWS Trainium/Inferentia, Google TPU, Microsoft MAIA/Cobalt)	Cloud providers become GPU substitutes and new chip vendors
Cisco	Amazon AWS (cloud networking), Arista, large enterprises with internal networks	Networking gear for data centers	Hyperscalers built their own switches and disaggregated network OS	Displaces traditional Cisco purchases with in-house designs
Oracle, Microsoft	Salesforce, Workday, ServiceNow	Databases and infrastructure for enterprise apps	SaaS firms built full-stack platforms competing with traditional enterprise software	Customers became full software suite competitors
Google Maps API	Uber, Lyft	Location services, navigation APIs	Ride-hailing firms built proprietary mapping to reduce dependence	Competes with mapping providers and reduces reliance on Google
Android/Google	Samsung (Tizen), Huawei (HarmonyOS)	Android mobile OS	Developed alternative smartphone OS platforms	Competing mobile ecosystems reducing Android dependency
AWS Marketplace vendors	AWS (Basics, managed services)	AWS acted as infrastructure + reseller of partner products	AWS launched services competing directly with partners (e.g., ElasticSearch/Opensearch, Datadog-like monitoring)	High-profile “customer-turned-competitor” ecosystem conflict
IBM, Dell, HP infrastructure	Major banks, retailers, healthcare systems	Enterprise servers, storage, and IT services	Internal cloud teams built private clouds replacing vendor systems	Vertical integration into infrastructure previously purchased
Facebook/Meta (mobile platforms reliance)	Meta’s VR/AR device program (Quest)	Reliance on Apple/Google mobile platforms	Meta developed its own hardware/software ecosystem	Competes with platform providers to escape dependency
Telcos buying vendor gear	AT&T, Verizon, Deutsche Telekom (open RAN initiatives)	Proprietary RAN equipment from Nokia/Ericsson	Built open-source or disaggregated RAN alternatives	Reduces dependence on traditional equipment vendors
IBM/Intel server vendors	Google, Amazon, Facebook data-center hardware	Commodity servers	Hyperscalers designed their own servers and power systems	Competing designs via OCP and private supply chain

That also can be seen in the market for neural processing units, where former customers Google and Amazon now have emerged as important suppliers of NPUs used in place of graphics processor units, even if many of the use cases are internal to those firms. 

Companies such as Google (Tensor Processing Units) and Amazon (Inferentia/Trainium chips) primarily use their NPUs internally or sell access through their cloud services, obscuring any direct "retail" market share comparison.

NPU Segment / Use Case	Dominant Architecture / Product Type	Market Share Context & Key Vendors	Key Vendor Dominance / Market Share Notes
Data Center / Cloud AI (Training,  Inference)	GPU (for Training,  General-Purpose AI),  ASIC/Custom NPUs (for specific Inference)	This segment includes hyperscalers using hardware internally (like Google's TPUs) or for cloud-based services.	NVIDIA holds a dominant share (often cited as 90%+ for high-end AI training accelerators/GPUs, which are often grouped with NPUs). Google (TPU), Amazon (Trainium/Inferentia), and AMD (Instinct) are the primary competitors in the custom/dedicated space.
Edge Devices (Retail/B2C)	Integrated NPUs (AI-SoCs) and Dedicated Edge NPUs	This segment covers chips embedded in consumer products for on-device AI (smartphones, PCs, smart home, automotive).	Qualcomm (Snapdragon), Apple (A/M series chips), and Samsung (Exynos) dominate the smartphone/tablet space, which accounts for the largest application share (e.g., 37.6% of the total NPU market application in 2024). Intel (Core Ultra) is a major player in the PC NPU market.
Edge Market Share (Application)	Smartphones & Tablets	The largest single application area, driving the growth of retail NPU units.	Estimated 35% - 40% of the NPU market application share is in retail, for smartphones and tablets.
Data Center NPU Share (Product Type)	Data Center NPUs	Market share based on the volume of processing units deployed in large-scale data centers.	Data Center NPUs maintained an estimated 51.6% of the neural processor market share in 2024, and much of that represents internal consumption by Amazon and Google.

Google designs and uses TPUs for its own services including search, translate, Gemini AI, for example. But Google does make its TPUs available to Google Cloud Platform customers as a service, though it does not sell the chips.

AI Factories Tend to Rely on Lots of Natural Gas

The hyperscale high-performance “computing as a service” providers (AWS, Azure, Google, Meta) mostly purchase renewable energy and report high market-based renewable energy percentages, and also use local power grid suppliers that rely on a high share of natural-gas generation (especially in Texas and Northern Virginia). 

Company / AI factory	Company claim (market-based renewables)	Major data-center regions (grid natural-gas share)	Degree of access to natural gas (High / Medium / Low)	Sources
Amazon / AWS	100% of electricity consumed matched with renewable energy (2023–2024, market-based). (Amazon Sustainability)	Heavy footprint in Northern Virginia (PJM/VA — VA generation >50% gas) and Texas (ERCOT — large gas share). US grids overall ~40% gas. (U.S. Energy Information Administration)	High	AWS has major capacity in gas-heavy US grids (Northern VA, TX). Although AWS reports “100% matched” renewables market-wide, the local grid supply for many AI clusters is still gas-dominated, so AWS has strong practical access to gas power for fast capacity scaling. (Amazon Sustainability)
Microsoft / Azure (incl. Azure OpenAI / OpenAI workloads on Azure)	Microsoft reports procuring enough renewables to match 100% of global electricity consumption (market-based). (Microsoft CDN)	Large Azure presence in Northern Virginia, Texas, and other US regions (many gas-heavy grids). US/TX/VA grid gas shares high (see EIA). (U.S. Energy Information Administration)	High	Microsoft powers OpenAI workloads on Azure; Azure’s major data-center regions overlap with gas-heavy grids (so operational access to gas is high despite market-based renewable matching). (The Official Microsoft Blog)
Meta (Facebook / Meta AI data centers)	Meta reports matching much (or all) owned/operated data-center electricity with renewables in its accounting (market-based) while expanding new local projects. (Meta Sustainability)	Big builds in Texas (Coleman County and other TX sites) and Northern Virginia; TX and VA grids have large natural-gas shares. (pv magazine USA)	High	Meta is rapidly expanding AI capacity in TX/VA — both regions with heavy natural-gas generation — so operational access to gas generation is high even as Meta signs renewables locally/contractually. (pv magazine USA)
Google / Google Cloud	Google reports having matched 100% of annual electricity consumption with renewables for years (market-based) and publishes regional hourly carbon-free percentages. (blog.google)	Google’s footprint includes Midlothian, TX (gas-heavy) but also WA/OR (hydro), and other lower-gas grids — a more geographically mixed footprint. (RTO Insider)	Medium	Because Google’s data centers are more geographically diverse (some gas-heavy, some hydro/low-gas), its practical access to gas is medium overall. It also reports regional CFE (carbon-free energy) metrics to show hourly variation. (blog.google)
Oracle Cloud	Oracle claims high renewable coverage in some disclosures (Oracle reports strong renewable procurement claims). (Oracle)	Oracle’s cloud footprint is smaller than hyperscalers and more concentrated in commercial colocation markets (regional mixes vary). Many U.S. colo grids include substantial gas. (Oracle Blogs)	Medium–Low	Oracle’s absolute compute footprint is smaller and it emphasizes renewable procurement; depending on region, local grid gas exposure varies — overall less direct exposure than the largest hyperscalers. (Oracle)
AI firms that rent cloud capacity (OpenAI, Anthropic, Stability, etc.)	OpenAI: uses Azure (Microsoft) for most workloads; Anthropic & others use mixtures of Google Cloud / Azure / AWS (multi-cloud). (The Official Microsoft Blog)	Their practical gas access ≈ the cloud provider(s) they run on. If on Azure/AWS/Google in TX/VA, access is High/Medium as above. (The Official Microsoft Blog)	Varies (follows provider)	These AI model operators rarely own global data centers; they rely on hyperscalers. So their degree of access to gas ≈ the hosting cloud’s regional grid exposure. OpenAI on Azure → High by the table above. Anthropic’s deals (Google/Azure) mean varied exposure. (The Official Microsoft Blog)

The “neo cloud” providers tend to have medium to high levels of access to natural gas for power, though the emphasis on “renewable” sources might not be as high a priority. 

Sites like CoreWeave’s Project Horizon and Galaxy/Helios (West Texas) are virtually designed around proximity to natural-gas infrastructure. For frontier-scale AI (multi-hundreds of megawatts to gigawatt scale), this gives them very-high access to gas-powered electricity.

TeraWulf’s Lake Mariner (NY) is a contrasting model: using mostly low-carbon grid supply (hydro/nuclear/clean energy) that is better for carbon-intensity, with lower reliance on gas.

Hut 8 has a mixed strategy: some sites (Canada) on cleaner grids, some (Texas, Panhandle) on gas-heavy or mixed grids, giving it a balanced, diversified exposure depending on where its compute load is run.

Riot Platforms (Rockdale / Corsicana) are among the most gas-exposed of publicly traded neo-cloud specialist compute providers. 

Company	Major sites / where compute is (or is planned to be) sited	Dominant local power sources / relevant company energy plans	Degree of access to natural gas power (High / Medium / Low)	Why / notes & primary sources
CoreWeave	Rapid expansion into West Texas / Permian projects (Project Horizon / Poolside JV), plus European builds (UK, Spain) and colo deals.	West Texas projects explicitly target locations with access to low-cost natural gas (Permian/Delaware basin) while some EU/UK sites emphasize renewable-backed supply. Net: strong access to local natural gas where it matters for large-scale AI campuses.	High	CoreWeave is anchoring large West Texas campuses that are being designed around low-cost gas-rich markets (Permian Basin/Project Horizon) while also deploying renewables-backed facilities in Europe. This gives CoreWeave high practical access to gas power for large-scale, fast-growing AI capacity. (Barron's)
Hut 8 Corp.	Mixed footprint: multiple Canadian colo/HPC sites (Vancouver, Kelowna, Mississauga/Vaughan), mining campuses (Alberta, Medicine Hat), and a growing U.S. development pipeline (planned US sites, Louisiana/Baton Rouge acquisitions announced).	Canada: strong hydro in some provinces (Quebec/BC/ON) → low gas; Alberta sites and some U.S. development pipeline → higher fossil/gas exposure. Company runs both renewables/hydro-backed sites and gas-exposed mining/power projects.	Medium	Hut 8’s compute estate is geographically mixed: many Canadian colo/HPC sites sit on low-carbon hydro grids, but Hut 8 also owns/operates power-first mining/data campuses in Alberta/other U.S. projects that expose it to gas/thermal generation. Net = medium. (Hut 8 HPC)
TeraWulf (TeraWulf Inc.)	Lake Mariner campus (Somerset/Buffalo, NY) — large hydro/low-carbon facility; Nautilus in Pennsylvania; plus announced / planned large campus development in Abernathy, Texas (joint venture with Fluidstack / Google backing).	Lake Mariner: predominantly hydro / low-carbon (NY grid + hydropower). Abernathy (TX) project would sit in gas-heavy Texas grids. Company messaging emphasizes “sustainably powered” HPC but also expansion into Texas.	Medium	TeraWulf’s core existing operations (Lake Mariner) are strongly low-carbon/hydro-aligned → low local gas exposure today. But an explicit expansion into Abernathy, Texas (large planned capacity) points to future higher gas exposure at those sites. Overall practical access = Medium (mixed existing low-gas + planned gas-region capacity). (Data Center Map)
MARA / Marathon Digital (MARA Holdings)	Large portfolio: Garden City, Granbury, McCamey (TX), multiple West Texas / Delaware Basin projects, plus international sites. Company explicitly pursuing integrated power + data center builds in West Texas (MPLX partnership).	Mix today: some wind/hydro-adjacent sites (Garden City adjacent to wind), BUT public plans to build gas-fired generation facilities in West Texas (MPLX partnership) — initial ~400 MW with potential to scale to 1.5 GW using natural gas from Delaware Basin.	High (increasing)	Marathon/MARA has both renewables-adjacent assets (e.g., Garden City wind) and explicit, recent plans/partnerships to build gas-fired generation co-located with data centers in West Texas (MPLX deal). That makes MARA’s practical access to natural gas high and rising as gas-gen projects come online. (Mara)
Riot Platforms (Riot)	Large Texas footprint (Rockdale, Corsicana) plus Kentucky facility — Rockdale is one of North America’s largest mining campuses.	Texas grids (ERCOT/North Texas) have very high shares of natural-gas generation at times; Riot’s large Texas facilities operate in gas-dominated grids and have been criticized for high fossil generation intensity. Riot’s filings list Texas facilities as core.	High	Riot’s major capacity (Rockdale, Corsicana) sits in Texas, a grid and market with large natural-gas generation share — giving Riot high practical access to gas-fired electricity for large scale compute/mining workloads. (Riot Platforms)

Of course, companies operating multiple sites will have some sites using more or less natural gas, depending on what other sources are available (hydro, for example). 

Company	Site	City	State	Natural Gas Access	Source
CoreWeave	Project Horizon (Longfellow Ranch)	Pecos County	TX	High	https://poolside.ai/blog/announcing-project-horizon; https://datacenterdynamics.com/en/news/ai-startup-poolside-teams-up-with-coreweave-on-2gw-data-center-in-texas/
CoreWeave	Helios (Galaxy) - Dickens County	Afton/Dickens County	TX	High	https://investor.galaxy.com/news/; https://www.datacentermap.com/usa/texas/dickens/helios-data-center/
CoreWeave	Livingston / NJ operations (representative)	Livingston	NJ	Medium	https://www.coreweave.com/news/coreweave-announces-partnership-with-foundation-model-company-poolside-to-deliver-ai-cloud-services
TeraWulf	Lake Mariner	Barker (Lake Mariner)	NY	Low/Medium	https://www.terawulf.com/lake-mariner-mining/; https://www.gem.wiki/Lake_Mariner_facility
TeraWulf	Nautilus (Pennsylvania)	Pittsburgh area (Nautilus)	PA	Medium	https://investors.terawulf.com/news-events/press-releases/detail/83/terawulf-announces-july-2024-production-and-operations
Hut 8	King Mountain / McCamey (JV)	McCamey	TX	Medium	https://hut8.com/2025/02/04/hut-8-operations-update-for-january-2025/; https://www.nasdaq.com/press-release/hut-8-operations-update-february-2025-2025-03-06
Hut 8	Vega / Texas-Panhandle (planned)	Vega	TX	Medium	https://hut8.com/news-insights/press-releases/hut-8-announces-plans-to-develop-four-new-sites
Marathon (MARA)	Garden City	Garden City	TX	Medium/High	https://baxtel.com/data-center/marathon-digital-garden-city-tx; https://www.mara.com/posts/mara-announces-25-megawatt-micro-data-center-project-powered-by-excess-natural-gas-from-oilfields
Marathon (MARA)	McCamey / West Texas projects	McCamey	TX	High	https://www.mara.com/posts/mara-announces-25-megawatt-micro-data-center-project-powered-by-excess-natural-gas-from-oilfields; https://ir.mplx.com/CorporateProfile/press-releases/news-release/2025/MPLX-and-MARA-Announce-Collaboration-on-Integrated-Power-Generation-and-Data-Center-Campuses-in-West-Texas
Riot Platforms	Rockdale	Rockdale	TX	High	https://www.riotplatforms.com/bitcoin-mining/rockdale/
Riot Platforms	Corsicana	Corsicana	TX	High	https://www.riotplatforms.com/bitcoin-mining/corsicana/; https://www.mapquest.com/us/texas/riot-platforms-inc-721940928
Core Scientific	Dalton (GA)	Dalton	GA	Medium	https://www.datacentermap.com/c/core-scientific/
Core Scientific	Grand Forks (ND)	Grand Forks	ND	Low/Medium	https://www.datacentermap.com/c/core-scientific/
Core Scientific	Muskogee (OK)	Muskogee	OK	Medium/High	https://www.datacentermap.com/c/core-scientific/
Compute North (historical)	Big Spring (TX)	Big Spring	TX	High	https://dgtlinfra.com/compute-north-chapter-11-bankruptcy-filing/; historical filings
Compute North	North Sioux City (SD)	North Sioux City	SD	Medium	https://dgtlinfra.com/compute-north-chapter-11-bankruptcy-filing/
Compute North	Kearney (NE)	Kearney	NE	Medium	https://dgtlinfra.com/compute-north-chapter-11-bankruptcy-filing/
TerraWulf (Beowulf/TeraWulf)	Lake Mariner (alternate coord)	Barker/Buffalo area	NY	Low/Medium	https://www.datacentermap.com/usa/new-york/buffalo/lake-mariner-data/; https://www.gem.wiki/Lake_Mariner_facility
TeraWulf	Nautilus (PA)	Pittsburgh area	PA	Medium	https://investors.terawulf.com/news-events/press-releases/detail/83/terawulf-announces-july-2024-production-and-operations
Greenidge	Dresden (Finger Lakes)	Dresden	NY	Low/Medium	https://www.greenidge.com/operations/
CoreWeave	Midlothian / Dallas area (representative)	Midlothian	TX	High	news coverage of cloud builds in Midlothian/Dallas area
Marathon (MARA)	Granbury (Wolf Hollow / Granbury)	Granbury	TX	High	Compute North and Marathon filings; company press releases