Sharon AI's $373M Deal Puts 120MW on a Lease Path-But the Real Bet Starts in 2027

The $373 Million Headline Matters, but Revenue Timing Matters More

Sharon AI has secured a five-year cloud computing service agreement worth US$373 million, but the financial payoff is still ahead. The company said revenue under the agreement is expected to commence during the first quarter of 2027, so investors are not getting an immediate earnings boost. Instead, the deal matters because it shows existing demand and suggests that Sharon AI's infrastructure buildout has a customer pathway.

Why the contracted capacity is the key near-term signal

For current holders, the more important point is utilization. After this agreement, Sharon AI has 120MW contracted to end customers out of 132 megawatts of total AI Factory capacity. That does not guarantee timing or margins, but it does suggest that most of the planned capacity already has a potential buyer attached. The main risk is no longer whether demand exists; it is whether that demand converts into deployed compute and billed revenue on schedule.

Why 120MW Contracted Changes How to View the Story

This is less about one contract and more about whether Sharon AI can repeatedly turn power capacity into deployed AI infrastructure and then into recurring revenue.

The mechanism: demand first, then capacity conversion

The company's model is becoming clearer: secure customer commitments, then convert megawatts and rack capacity into live compute. The Australia project is a useful example, with 72MW of AI data center capacity tied to NVIDIA's DSX design and a route to 40,000 Grace Blackwell GPUs. The latest agreement adds another concrete milestone, with Sharon AI expecting to upgrade from 62,000 to 64,000 NVIDIA GPUs across its AI Factory platform by mid-2027. The initial deployment under this agreement is expected to use 2,048 NVIDIA Blackwell Ultra B300 GPUs, giving investors a tangible starting point for tracking deployment progress.

Why the balance between capacity and deployment matters

The numbers matter because they show where the business is headed. Sharon AI is no longer just talking about infrastructure; it is linking capacity, GPU deployment targets, and customer commitments. That shifts the debate from "does anyone want capacity?" to "how quickly can unused megawatts become monetized compute?" If deployment keeps pace with announced capacity, investors will have a stronger case for treating Sharon AI as operating infrastructure rather than only a future narrative.

The main risk is execution, not demand

The bear case is straightforward: with 120MW contracted to end customers out of 132 megawatts of total capacity, the easy part may already be behind the company. Future upside now depends more on deployment speed, equipment supply, and whether Sharon AI can keep timing targets intact. If the final stretch of capacity converts after the expected 2027 revenue start, the stock has a clearer path to being judged on operating execution rather than on future potential alone.

What to Watch as the Buildup Continues

For now, this looks like a buildup story, not a cash-flow story. The next hard checkpoint is revenue commencing during the first quarter of 2027 under the latest five-year agreement. If that date holds and deployment follows, investors can start looking for signs of operating leverage. If it slips, the shares are more likely to remain a bet on future conversion rather than evidence of mature monetization.

Signals that would strengthen the thesis

• Evidence that Sharon AI is still moving toward 64,000 NVIDIA GPUs across its AI Factory platform by mid-2027.
• Signs of deployments beyond the expected initial deployment under this agreement is expected to utilize 2,048 NVIDIA Blackwell Ultra B300 GPUs.
• Updates that tie commercial activity more closely to the expected first quarter of 2027 revenue start.
• Any change in the balance between total capacity and the 120MW contracted to end customers.

Signals that would weaken it

• The first quarter of 2027 revenue start slips without a credible replacement timeline.
• Progress toward 64,000 NVIDIA GPUs across its AI Factory platform by mid-2027 stalls.
• There is little visible progress beyond the expected 2,048 NVIDIA Blackwell Ultra B300 GPUs.
• The gap between total capacity and the 120MW contracted to end customers stops narrowing, suggesting fewer near-term conversions.