Meta has done something this week that we haven't seen from the company in quite a while.
For some time, Zuckerberg's company had been lagging behind in AI. Its models never reached the level of OpenAI, Anthropic, or Google. Then, within just a few days, Meta announced three major moves at once: a new AI model, its own AI chip, and a multibillion dollar data center in Canada.
What does this mean? Meta is finally making an aggressive push into the AI race and trying to catch up with its biggest competitors.
THE NEW MODEL
Let's start with the first, and arguably the most important announcement. Meta introduced Muse Spark 1.1, a model that the company says is a major upgrade over its predecessor.
What exactly does it do? It is designed for "agentic" AI, meaning systems that don't just answer questions but can independently complete complex tasks. It offers major improvements in coding, tool use, and can retain an enormous amount of information thanks to a context window of one million tokens.
Here's where things get interesting. Zuckerberg himself said the new model outperforms Google's Gemini 3.1 Pro.
"This may be the first time Meta's models are better than all of Google's models."
But the biggest change is somewhere else. For the first time, Meta is moving away from open source for one of its flagship models. Instead, it is keeping the model closed and selling access through an API.
"It's the first time we're offering a truly serious API," Zuckerberg said.
The pricing is aggressive as well. The API costs roughly one quarter of what OpenAI and Anthropic charge, and every new account starts with $20 in free credits. The strategy is clear: encourage as many developers as possible to build on the platform.
Zuckerberg was so excited about the launch that he even posted about it on X for the first time in three years.
ITS OWN AI CHIP
The second major announcement focuses on hardware.
Until now, Meta relied on chips from Nvidia and AMD to power its AI infrastructure. That is about to change.
The company's first in house AI chip, called Iris, is scheduled to enter production in September.
Meta is developing the chip with Broadcom handling the design and TSMC manufacturing it. Even more impressive, internal testing reportedly lasted only six weeks without any major issues.
Why is Meta doing this?
The goal is to reduce dependence on outside suppliers while dramatically increasing computing capacity. The company aims to reach 14 gigawatts of AI infrastructure next year, up from roughly 7 gigawatts this year, effectively doubling its computing power in just twelve months.
Meta has also secured long term agreements with Samsung for memory, Sandisk for storage, and Sumitomo Electric for optical fiber.
The message is clear: infrastructure, infrastructure, infrastructure.
A NEW DATA CENTER IN CANADA
The third announcement is just as significant.
Meta has begun construction of its first Canadian data center in Sturgeon County, Alberta.
This is not an ordinary facility.
The investment exceeds 13 billion Canadian dollars and will become Meta's 33rd data center worldwide. The project is expected to create around 3,000 construction jobs and more than 300 permanent positions once completed.
The facility is being purpose built for AI workloads. It will run entirely on renewable energy and use a closed loop cooling system designed to minimize water consumption.
Why spend so much money?
Because AI requires enormous computing power. Any company that wants to compete at the highest level must build massive infrastructure, and Meta is investing at an extraordinary pace to make that happen.
WHAT ANALYSTS THINK
The most important question is how Wall Street views all of this.
There is still a major concern.
For months, investors have been pressuring Meta to prove that these massive AI investments will eventually generate meaningful returns. Back in April, the stock fell nearly 9 percent after earnings, despite reporting strong revenue and profit. Investors were worried that the company still lacked a clear plan to monetize its enormous AI spending.
On the other hand, there is also a more optimistic view.
Needham analyst Laura Martin believes Meta's shift toward proprietary closed models represents a major opportunity. She points out that Muse Image is already powering Instagram, WhatsApp, and the company's advertising tools.
The logic is simple.
Better AI generated images lead to better performing advertisements, which ultimately improves advertising revenue.
As Martin put it, Meta is becoming "a winner from AI rather than being replaced by it."
