Meta (META) stock climbed higher on Friday, as the company revealed a handful of potential new revenue streams that could help offset its prodigious spending on its AI buildout.
Shares of the social media giant rose more than 5% as of midday, turning the stock positive year to date. Meta’s stock was down year to date on concerns about rising capital expenditures and questions about returns on its massive AI investments.
On Thursday, CEO Mark Zuckerberg told Bloomberg in an interview that the company is exploring renting its own AI computing power to third parties.
While Zuckerberg didn’t say how exactly Meta would sell its data center capacity, it could either serve up and power AI models from competitors to its customers, like Amazon’s (AMZ) AWS, or simply sell access to its AI chips and servers like a neocloud provider. Meta announced on Wednesday that it will build a new data center in Canada, its 33rd such facility.
AI companies have repeatedly told shareholders that they are resource constrained on AI, and need more computing capacity to keep up with customer demand. If Meta grabs a slice of the cloud computing market, it could provide a hot, new revenue stream outside of its advertising business.
In addition to the potential data center business, Meta on Thursday announced its Muse Spark 1.1 AI model, as well as a pricing structure for how it will give developers access to the software that dramatically undercuts competitors like Anthropic (ANTH.PVT) and OpenAI (OPAI.PVT) on the cost of input and output tokens.
Rolling out a new, capable AI model would normally provide a vote of confidence for Meta’s Meta Superintelligence Labs efforts on its own, but its pricing is the bigger story.
Tokens are a unit of measurement in AI models. Each token generally represents pieces of words or phrases. Input tokens are the questions you ask a chatbot or commands you give an AI agent. Outputs are the responses a model spits out.
Meta says it will charge developers $1.25 per million input tokens and $4.25 per million output tokens. Anthropic currently charges $5 per million input tokens for the model and $25 per million output tokens for its Opus 4.8 model, which Meta compares with Spark 1.1.
In a post-tokenmaxxing world, Meta’s less expensive Spark 1.1 could attract price conscious developers who don’t need the raw power of something like Anthropic or OpenAI’s most powerful models.
