News
Meta to Double AI Compute by 2027, Builds Own Chip and $13 Billion Alberta Data Center
A leaked internal memo reveals Meta plans to double its total compute capacity to 14 gigawatts by 2027, launching its own AI chip code-named Iris and a massive $13 billion data center in Alberta, Canada.
Contents
Meta has unveiled plans for a dramatic expansion of its own AI computing infrastructure. The company intends to double its total compute capacity to 14 gigawatts by 2027, begin production of its own AI chip as early as September, and build its first data center in Canada at a cost of more than $13 billion.
News of Meta's plans leaked through an internal memo obtained by Reuters. The document frames the company's strategy as an attempt to build a third computing power alongside Amazon Web Services and Microsoft Azure, one based on its own hardware rather than relying solely on chip purchases from Nvidia.
Its own chip to cut Nvidia reliance
The chip, code-named Iris, designed in partnership with Broadcom and manufactured by Taiwan's TSMC, is set to enter production as early as September 2026. Meta plans to roll out a new generation of its in-house chips roughly every six months, a pace considerably faster than the semiconductor industry's standard cycle.
The company has also signed long-term supply agreements with SanDisk, Samsung Electronics and Sumitomo Electric, covering flash memory, memory chips and fiber-optic equipment respectively. These contracts are meant to secure the supply chain for components needed to build data centers at this scale.
A data center in the heart of Alberta
The new data center will be built in Sturgeon County, northeast of Edmonton in the Canadian province of Alberta. The investment, worth more than 13 billion Canadian dollars, will initially draw on 1 gigawatt of power, with the option to expand to 1.8 gigawatts. Electricity will come from a dedicated gas-fired power plant built by a joint venture between Pembina Pipeline and Kineticor Asset Management.
We believe a data center can only succeed if the community succeeds alongside it, and we want Sturgeon County to thrive - Gary Demasi, Vice President of Data Center Strategy, Meta
Alberta Premier Danielle Smith pointed to the investment as evidence of the province's growing importance in the global race for AI infrastructure. Sturgeon County Mayor Alanna Hnatiw, meanwhile, stressed that the region is becoming part of an emerging Canadian energy-and-compute corridor, bringing both economic opportunity and new responsibilities.
Market reaction and doubts
News of the chip plans and infrastructure buildout sent Meta shares up several percent in a single day, and over the week the stock posted its biggest gain since February 2024, adding roughly $13 billion to Mark Zuckerberg's net worth. Investor enthusiasm cooled somewhat after leaked excerpts from an internal company meeting revealed that Zuckerberg had admitted Meta's AI agent development had not accelerated as fast as the company had expected.
That gap highlights the tension between two tracks of Meta's strategy: aggressively building out hardware infrastructure on one hand, while facing uncertainty over the pace of real progress in its AI models and products on the other. Investors are watching ever more closely whether tens of billions of dollars in spending are translating into proportional product progress.
Implications for the global AI market
Meta's decision to build its own chips fits a broader trend among the largest tech companies seeking to reduce their dependence on Nvidia and lower the unit cost of compute. Google is already pursuing a similar path with its TPU chips, as is Amazon with its Trainium chips, signaling growing competition in the market for specialized AI processors.
For Polish companies using cloud services and AI models hosted by the big platforms, this could mean greater availability of compute capacity in the medium term, along with possible downward pressure on inference pricing once infrastructure supply starts catching up with demand. The scale of the investment also shows just how capital-intensive the race for AI advantage has become.
Sources: Reuters via Euronext (live.euronext.com), CNBC (cnbc.com), The Globe and Mail (theglobeandmail.com), CBC News (cbc.ca), Global News (globalnews.ca), BigGo Finance (finance.biggo.com)
