Core CPI accelerated on a monthly basis for the second consecutive reading. 0.2% to 0.3%. That is the number the market is underweighting.
The headline print landed at 3.4% year-over-year โ flat, again. Core CPI fell to 2.4% from 2.5%, which sounds like progress. It is not. The monthly annualized rate of core inflation is now running near 3.6%, more than double the Federal Reserve's target. Noise in the yearly comparison is masking a genuine re-acceleration.
CICC Research published its read on the August CPI data on September 12, concluding that the Federal Reserve has reached the threshold for a 25-basis-point hike at the September 16 FOMC meeting. The market, fixated on the near-term decision, largely agrees. The terminal rate path โ the 2027 and 2028 dot plot projections โ is where the real repricing risk sits.
Speed is the only currency that never depreciates. But in this cycle, duration is the exposure that never stops compounding.
The Threshold Language Game
CICC deliberately used the phrase "raise threshold" rather than "will raise." That is not hedging. It is a precision signal from a sell-side desk that understands how forward guidance operates in a data-dependent regime.
Translation: the Fed has sufficient evidence to justify a hike, but retains the flexibility to pivot if the data breaks. The market has already priced the September move at roughly 80% probability. The asymmetry is not in the September decision. It is in what the Summary of Economic Projections reveals about the long end.
Here is what most commentary misses: CICC flagged that the 2027 and 2028 dots are likely to be revised upward. That is a statement about the neutral rate, not about the next meeting. If the median dot for 2027 moves above 3.75%, the entire yield curve needs to reprice. Not just the front end โ the whole curve.
Based on my surveillance work during the 2024 ETF approval window, I learned that the market consistently underestimates how slowly central bankers revise structural assumptions. The dots move in increments. But when they move, the adjustment is vicious.
The AI Inflation Factor Nobody Is Modeling
This is the second time in 2025 that a major sell-side house has explicitly introduced "AI-driven inflation pressure" into a policy analysis framework. CICC's inclusion of it is not cosmetic. It is a structural break in how inflation dynamics are modeled.
Traditional inflation frameworks rely on two drivers: demand-pull and cost-push. The new framework requires a third leg โ supply constraints created by technology infrastructure buildout. AI data centers consume electricity at unprecedented rates. They require specialized cooling systems, advanced semiconductors, and grid upgrades. All of these inputs are inelastic in the short run. That creates a floor under producer prices that no demand slowdown can remove.
CICC specifically cited telecom price increases as a contributor to core services inflation. In any other cycle, telecom would be a footnote. In this cycle, it is a canary. 5G expansion and AI-driven data transmission demand are pushing communication costs higher. This is not rent. It is not medical care. It is the physical infrastructure of the AI economy transmitting directly into the CPI basket.
The Fed is watching this. The market is not.
Why the Monthly Rate Trumps the Yearly Rate
The annual core CPI number has declined for four consecutive months. That is the number policymakers cite in public statements. The monthly number โ the one that actually drives policy decisions โ has accelerated.
August's 0.3% monthly core gain, following July's 0.2%, translates to a 3.6% annualized pace. That is not transitory. That is not base effects. That is momentum.
CICC's read is that the Fed has begun pivoting to a monthly-annualized framework. This matters because it changes the threshold for action. A monthly core print of 0.3% is tolerable under a yearly-target framework. Under a momentum-based framework, it triggers intervention.
The 2025 policy framework is no longer "data-dependent." It has evolved into "inflation-persistence dependent." The Fed is no longer asking whether inflation is falling. It is asking whether the trend rate of disinflation is sufficient to return to target within a reasonable horizon. The answer, based on August data, is no.
Unemployment projections being revised lower reinforces the same conclusion. A tight labor market with sticky core services inflation means the Phillips curve โ declared dead in the early 2020s โ is twitching again. The Fed has room to tighten because the economy is not breaking.
Resilience is built in the quiet before the crash. The market's quiet confidence in a soft landing is precisely the condition that allows the Fed to keep tightening.
The Bear Steepening Trade Nobody Has Positioned For
If the dots shift higher for 2027-2028, the immediate consequence is a bear steepening of the Treasury curve. Long-end yields rise faster than short-end yields because the market is forced to revise its estimate of the average funds rate over the next three years.
The 2s10s spread has been range-bound for months. That range breaks on September 16 if the dots confirm CICC's read. A break above 4.5% on the 10-year is the psychological threshold that triggers systematic selling from duration-constrained funds.
Here is the cross-asset transmission chain that institutional desks should be mapping right now:
- U.S. Treasury duration risk reprices first, because it is the reference asset for global risk premia.
- The dollar strengthens as yield differentials widen. DXY above 105 is the trigger level for emerging market stress.
- USDCNH pushes beyond 7.30, forcing the PBOC to tighten conditions or deplete reserves defending the fix.
- China's monetary easing space compresses, pressuring A-share liquidity and Hong Kong equity valuations.
- Growth stocks in the U.S. face DCF discount rate pressure, while financials benefit from net interest margin expansion.
That is the playbook. It is not speculative. It is mechanical.
The edge lies in the data others ignore. The 0.3% monthly core print is ignored because the yearly number still shows disinflation. The 2027 dots are ignored because the September hike is already priced. The AI inflation factor is ignored because it does not fit the old framework. All three ignored signals point in the same direction: higher rates for longer.
The Fiscal Trap Hidden Inside the Rate Story
Here is the angle almost no coverage is touching. CICC's report is about monetary policy, but the structural consequences are fiscal.
The United States is running a deficit above 6% of GDP. Interest expense on the federal debt has already exceeded defense spending. Every rate hike raises the cost of new issuance. Higher rates lead to larger deficits. Larger deficits require more issuance. More issuance needs higher yields to attract buyers. Higher yields lead to... more hikes.
This is the reflexive feedback loop embedded in "higher for longer." It is the 2023 August meltdown scenario scaled up: Treasury auctions drawing tepid demand, yields spiking on supply, and the Fed forced to watch while its own tightening amplifies fiscal fragility.
The indirect bid ratio at Treasury auctions is the metric to monitor. Below 70% signals foreign demand is waning. Combined with the Fed's quantitative tightening continuing, that is the liquidity cocktail that breaks markets โ not the September hike itself.
Chaos is just data waiting for a pattern. The pattern in this cycle is tightening that cannot stop because inflation is structurally embedded in the AI buildout, and fiscal policy cannot help because debt service consumes an ever-larger share of revenue.
The Apple Scenario and Its Opposite
There is a non-zero probability that CICC is wrong. If the September SEP cuts the 2025 and 2026 rate paths โ a dovish surprise โ the entire bear steepening thesis unwinds in 48 hours. That is the tail risk that undermines the short-duration positioning most macro funds are running.
But the asymmetry favors the hawkish scenario. Core monthly inflation is accelerating. Labor markets are tight. AI capex is accelerating, not decelerating. Treasury supply is growing. Every observable data point pushes in one direction.
The more dangerous scenario is the 1970s replay, where AI investment collapses just as energy prices spike. That is stagflation with a technological twist โ inflation expectations break anchor, and the Fed faces an impossible choice between fighting prices and saving growth. Probability is low. Consequence is catastrophic.
Position for the September 16 Repricing
The next 72 hours determine the macro regime for Q4 2025.
The first signal is the dot plot at 2:00 PM ET on September 16. Median 2027 dot above 3.75% is hawkish. Median 2028 dot above 3.5% is a structural shift. The second signal is the Chairman's press conference. Phrases like "persistent" or "insufficient progress" confirm the persistence framework.
Then the data deluge begins. September CPI on October 10. Non-farm payrolls in early October. The PBOC's Q3 monetary policy committee statement at the end of September. Treasury's quarterly refunding announcement in late October. Every one of these is a potential repricing event.
The single most important thing to understand: the September hike is the distraction. The real signal is the long-duration dot plot. The market's conviction that "higher for longer" is fully priced is the most dangerous consensus in global markets right now.
Watch the 10-year at 4.5%. Watch DXY at 105. Watch USDCNH at 7.30. Watch the indirect bid ratio.
One of those breaks first. The others follow.
The question is not whether the Fed hikes on September 16. The question is whether the market has priced the Fed that refuses to stop hiking in 2027. Based on current positioning, it has not.
When that repricing hits, it will not be gradual.
Speed is the only currency that never depreciates. But in this cycle, duration is the exposure that never stops compounding.
