Новая инфляционная переменная ФРС: инвестиции в ИИ поднимают нижнюю границу затрат

Invest
Dark economic diagram showing AI infrastructure, power costs, semiconductor prices, and the US rate path lifting the operating cost floor
The June FOMC minutes suggest that waiting for rate relief is less useful than mapping AI-driven cost pressure, inflation expectations, and funding risk.

Практический сигнал не в ожидании снижения ставок. Спрос на ИИ-инфраструктуру может повышать цены на электроэнергию, чипы, облако и оборудование раньше, чем производительность улучшит маржу.

Confirmed facts: the vote was unanimous, but the cost map widened

  • The Fed kept the federal funds target range at 3.50%-3.75% on June 17, 2026, and the decision was approved by a 12-0 vote.
  • The minutes released on July 8 cited April PCE inflation of 3.8% and core PCE inflation of 3.3%. Based on CPI and PPI data available at the meeting, staff estimated May PCE at 4.1% and core PCE at 3.4%; BEA’s May pages now show those same year-over-year readings.
  • The BLS May CPI report showed headline CPI at 4.2% year over year and the energy index up 23.5%. The June CPI release is scheduled for 2026-07-14.
  • The June SEP moved the 2026 median federal funds rate projection to 3.8%, the 2026 PCE inflation median to 3.6%, and the core PCE median to 3.3%.
  • The New York Fed’s June Survey of Consumer Expectations showed one-year inflation expectations at 3.7%, three-year expectations at 3.3%, and five-year expectations unchanged at 3.0%.
  • The U.S. Treasury daily curve showed the 2-year yield at 4.21% and the 10-year yield at 4.56% on July 8, 2026.
Fed minutes operating map
SignalLatestOperating read
Fed target range3.50%-3.75%Policy relief has not arrived
2026 SEP funds median3.8%Higher than the current midpoint
May PCE / core PCE4.1% / 3.4%The Fed’s preferred inflation gauge is still high
NY Fed 1Y / 3Y expectations3.7% / 3.3%Expectations risk is not only an energy story
10Y Treasury, Jul 84.56%Long rates keep funding costs elevated

Interpretation: AI is a cost story before it is a productivity story

The headline fact is simple: the Fed held rates steady. The operating signal is more complicated. Officials were not only debating tariffs, energy, and Middle East supply risk. The minutes also treated AI infrastructure demand as a source of upward pressure on technology products and electricity.

That matters because AI tools are not only productivity assets. They are also invoices. API calls, GPU instances, cloud commitments, data-center power, high-performance memory, networking equipment, and engineering time hit cash flow before the productivity gain is easy to measure.

Investors need the same separation. AI capex can lift revenue for chip, power, cloud, and infrastructure companies. The same boom can also keep discount rates elevated if it supports inflation and pushes expected policy rates higher. Strong earnings and a higher cost of capital can exist at the same time.

The minutes also underline a credit split. Financing remained generally available for larger businesses, while conditions were tighter for many small businesses and lower-score households. AI bond issuance by large companies does not make a small team’s working-capital line cheaper.

Market narrative: strong AI earnings and sticky inflation can coexist

The market narrative is split between a long-run productivity story and a near-term cost story. The optimistic version says AI adoption eventually expands supply and reduces unit costs. The harder version says electricity, chips, equipment, rent, and labor can rise before that productivity benefit shows up.

The minutes are important because they keep both ideas alive while separating the timing. Productivity may arrive later; invoices arrive this quarter. Operators should not respond by cutting every AI tool. They should concentrate usage where ROI is visible and cap experimentation elsewhere.

Rates follow the same logic. Waiting for lower rates before repairing the cost structure is a weak plan. The June SEP moved the median 2026 policy rate higher than in March, while short- and medium-term inflation expectations rose. Lower-rate assumptions should not be the default operating case.

Second-order effects small teams may feel first

AI tool budgetsExperimental API calls, duplicate SaaS seats, and idle credits become cash leakage when rates stay high.
Power and infrastructure costsData-center and high-performance computing demand can feed later into cloud pricing and contract terms.
B2B pricingCustomers face the same pressure. Price increases work better when tied to usage, automation savings, or lower support costs.
PortfoliosAI revenue growth must be weighed against free cash flow, leverage, power exposure, and valuation sensitivity to higher yields.

This week’s operating checklist

• Split AI and cloud spend into direct revenue contribution, internal productivity, and experimentation; put a monthly cap on experimentation.

• Put fixed-rate debt, floating-rate debt, card payment dates, and annual SaaS renewals into one cash-flow calendar.

• Test price changes tied to usage, AI features, or premium support before broad price increases.

• For investments, review free cash flow, net debt, power contracts, chip supply exposure, and data-center commitments alongside AI revenue growth.

• Track rates, FX, failed payments, and cloud usage around 2026-07-14 CPI, the 2026-07-28/29 FOMC meeting, and 2026-07-30 PCE.

Risks and counterarguments

The counterargument is real: AI adoption can eventually raise productivity and lower unit costs. The minutes also acknowledge that productivity gains could expand aggregate supply over time.

Inflation may cool if energy prices and Middle East supply risks ease. The New York Fed survey even showed lower gas-price expectations. It would be too strong to treat AI demand alone as a guaranteed rate-hike trigger.

The practical risk is timing. Small teams do not pay bills with eventual productivity. Until the gain is visible in margins, assume the cost floor is higher and manage runway accordingly.

Disclaimer: Материал носит информационный характер и не является финансовой рекомендацией.

Источники

다른 글