✨ Big check in

Big Tech is writing some very big checks for AI. Goldman Sachs estimates global AI infrastructure spending could surpass $1 trillion this year. That bill goes well beyond GPUs, covering the data centers, power systems, and other infrastructure needed to put all that compute to work. 

As AI clusters get larger, thousands of chips need faster links to move data between them, driving demand for switches, transceivers, lasers and fiber. This earnings season is giving investors an important reminder that orders, sales, and deployments don't necessarily move on the same timetable.

Take Cisco: The networking giant booked $9.3 billion of AI infrastructure orders from hyperscalers in fiscal 2026, including $4 billion in the fourth quarter alone according to its earnings report. Cisco had guided to roughly $4 billion of AI infrastructure revenue for fiscal 2026 and now expects that figure to reach $7.5 billion in fiscal 2027. Still, traders heavily focused on a decline in product gross margins, which dropped to 66.3% from 68.4% the previous year, and were not impressed by the revenue forecast. 

Cerebras Systemscan serve as another example. The chipmaker reported $180.1 million in second-quarter revenue in its earnings statement, up 74.3% from a year earlier, but below the $194.23 million analysts expected according to Reuters. Its cloud-services revenue roughly quadrupled to $126 million, while hardware sales fell to $54.1 million from $70.3 million a year earlier.

Don’t get me wrong, the AI infrastructure boom is still real and hot. Recent earnings show demand piling up across the AI stack:

• Coherent saw Datacenter & Communications revenue jump to $1.62 billion last quarter.

• Applied Optoelectronics reported revenue growth, with data-center revenue surging to $107.7 million.

• Nebius grew quarterly revenue to $582.3 million, with revenue tied to AI cloud business increased sixfold to $575 million. 

• Corning: Optical Communications sales rose 32% to $2.07 billion.

As demand for the AI buildout keeps surging, investors are starting to care more about how quickly the AI orders can be delivered, recognized as revenue, and converted into earnings.

The Takeaway

For much of the AI boom, proving demand was not hard. Huge capex plans, record orders and swelling backlogs gave investors a reason to bet that years of growth were still ahead. This earnings season suggests the bar for many of these stocks has gotten higher.

An order booked today can represent real future business, but it still takes time to manufacture the equipment, complete the deployment and recognize the sale. During that wait, costs can rise, margins can move, and the expectations already embedded in a stock can keep climbing.

For investors, that means not every giant AI backlog deserves to be valued the same way. The companies that can convert demand into revenue quickly while protecting margins may be rewarded very differently from those whose order books grow faster than their ability to deliver.

- Helena Cheng

- J. Edward Moreno

The streaming and cable deals to air “Seinfeld” have been renewed, with Netflix getting the rights to stream the show for another five years and Paramount getting the basic cable rights for another three. The sitcom remains one of the most popular shows, bar none; Nielsen says 60 million Americans have watched Seinfeld in the first six months of this year.

• July Retail Sales

• Preliminary August consumer sentiment