Sacks Calls 2028 Deadline: AI Data Center Pipeline May Close

David Sacks put a date on the end of the AI data center buildout: 2028. On the All-In podcast episode released November 7, he said: "Nobody can make a $50 billion investment in conditions of uncertainty. I think by 2028 the pipeline will be closed or significantly reduced."

His argument rests on permitting, not demand. Sacks warned that the biggest danger is breaking ground and then, halfway through construction, getting "enjoined" — stopped by a worker strike or a community group. He added that governors in the largest data center states are "basically indicating that they favor a moratorium" until certain conditions are met. A forecast, plainly. Proposed moratoriums can be narrowed, delayed, or dropped.

Why a Permitting Freeze Reaches Every Supplier

A data center needs land, power, and a finished building before a single chip arrives. Broadcom's chief executive made the same point on the company's September call: "Before you get the chips out there, you have to get the data center shell in place and ready for production." If the buildings stall, chip, networking, and cooling orders slow after them.

That dependency matters because every company in the chain is posting growth that assumes continued expansion:

• NVIDIA (NASDAQ:NVDA) reported Data Center revenue of $89.02 billion, up 117%, guided next quarter to $108.0 billion plus or minus 2%, and disclosed supply obligations of $279.00 billion in its latest SEC-filed earnings release.
• Broadcom (NASDAQ:AVGO) sees fiscal 2028 AI semiconductor revenue reaching $230 billion. That target falls squarely inside Sacks's 2028 window.
• Arista Networks (NYSE:ANET) grew revenue 58.5% and guided Q4 to $3.9–4.0 billion, while disclosing that a limited number of large customers drive much of that revenue.
• Vertiv (NYSE:VRT) reported organic orders up over 100%, with customers typically ordering 12 to 18 months ahead.

Landlords Held Up While Chip Stocks Fell

The market has already begun pricing the split. Over the month ending November 6, NVIDIA fell 21.4%, Broadcom 22%, Arista 20.86%, and Vertiv 25.81%. Over the same month, Equinix (NASDAQ:EQIX) slipped only 1.85% and Digital Realty 3.47%.

One explanation: a freeze could help companies that already own permitted, powered sites. Equinix's chief executive said: "If there are a number of projects canceled, then it only makes what is already scarce capacity much more of a scarce resource." She added that across its 3 gigawatts of planned capacity, "the land, power and permitting are largely secured." Digital Realty has 1.5 gigawatts under construction, 63% pre-leased, and a record $2.5 billion backlog that gives it visibility through 2027 and into 2028.

What Is the Counterargument?

The optimists are not quiet. Vertiv's chief executive called power, permitting, and labor pacing items that "are not slowing down growth." NVIDIA expects fiscal 2028 revenue growth of approximately 70% and describes that outlook as limited by supply, with customer demand running higher. The company also cited top-five hyperscaler capital spending of nearly $800 billion in 2026 and $1.3 trillion in 2027.

What Should Investors Watch Next?

Many people own this trade without picking a single stock. A major Nasdaq-100 index fund's prospectus notes significant exposure to information technology, so a slowdown in AI spending would surface in many 401(k) and IRA balances. Four signals stand out:

1. State action — whether moratorium talk in the big data center states turns into passed rules or stays at the proposal stage.
2. Guidance checks — NVIDIA's results against its $108.0 billion guide, and Broadcom's against expected Q4 AI revenue of $21.7 billion.
3. Vertiv orders — with a 12-to-18-month ordering lead time, an order slowdown would be among the first signs that 2028 projects are slipping.
4. REIT pre-leasing and pricing — rising rents at Equinix and Digital Realty would suggest scarcity is helping landlords even as new construction slows.

If permitting tightens, investors with heavy index exposure to AI hardware should distinguish which parts of the chain depend on new construction and which collect rent on capacity already built.