M2 Is Growing Into a 4.1% Inflation Rate

Buried in the Fed's semiannual report is a useful number: average M2 from January through May was 4.7% higher than a year earlier. The Fed says money growth has returned to something closer to its 2010s pace, while M2 velocity sits only slightly below its pre-pandemic level.

Headline PCE is already at 4.1%, and the money stock is expanding again. The Fed calls its stance restrictive. Savers should watch what the institution does, not what it calls itself.

It can move the price of money. It cannot print electricity, transformers, housing, or barrels of oil. Bitcoin was built around the assumption that this institution will keep failing in the same predictable direction.

Cerebras Reports $25 Billion of Remaining Performance Obligations. Read the Fine Print.

Andrew Feldman told the All-In podcast that AI demand is outrunning the industry's ability to build data centers and fill them with hardware. He described Cerebras as having a $25 billion backlog and said, "the demand is booked."

Cerebras's filing gives the claim substance, but the viral interpretation runs ahead of the details.

Cerebras's first-quarter 10-Q reports $25 billion of remaining performance obligations as of March 31. That accounting balance includes deferred revenue and amounts expected to be invoiced and recognized in future periods. A significant amount relates to the OpenAI master relationship agreement, and the balance also includes certain data-center rent, leasehold, security, power, and utility pass-through costs.

The company expects to recognize approximately 16% during the initial 24 months ending March 31, 2028, another 45% during months 25 through 48, and the remainder thereafter, although customer delivery timing may change that schedule.

This is contracted future revenue scheduled across several years. It is not $25 billion of chips shipping next quarter. Cerebras still has to turn those obligations into energized buildings, installed systems, working networks, and delivered compute. Software can scale with a keystroke. Power plants, substations, and transformers cannot.

The Bear Is Where the Next Bull Gets Built

James Check took a break from measuring capitulation to ask a more enjoyable question in Numba Go Up: where could the next bitcoin cycle peak land if the long-term floor models keep compounding?

He is explicit that this is speculation. Good. Models are tools, not promises.

In his paid July 10 post, Check presents speculative model outputs that put a future bear-market floor around $129,000 to $179,000. He then applies a 2x multiple to produce $258,000 to $358,000, while his vaulted-price and power-law models reach roughly $362,000 and $480,000, respectively. These are Check's model outputs, not independently verified forecasts. He puts his own central scenario north of $300,000.

Do not read that paragraph as a price target. Read it as a reminder of what bear markets do. They crush the euphoric inputs. They depress the growth measurements. They remove the tourists. They transfer coins from people who need excitement to people who understand the asset.

The next bull gets built while everyone is bored, angry, and convinced the number will never go up again.

Bitcoin Policy Institute Moves to Intervene in the Dormant-Wallet Fight

The Bitcoin Policy Institute has moved to intervene as a defendant in the Noah Doe dormant-wallet lawsuit in New York.

NYSCEF Document No. 65, filed July 10 under Index No. 153119/2026, is a 22-page memorandum supporting BPI's intervention in ABC Company, XYZ Company, and Noah Doe v. John Does 1-39,069. White & Case submitted the filing. The court has not granted intervention, and none of BPI's arguments should be confused with a ruling.

BPI says it holds bitcoin as a long-term reserve asset and intends for those coins to remain unmoved for long periods. It argues that treating visible on-chain dormancy as abandonment could expose its holdings, and other long-held UTXOs, to manufactured title disputes.

We covered the underlying case on July 3. The principle has not changed. Dormancy is not abandonment. A public address is not property someone found on the sidewalk. Refusing to move your coins is an exercise of ownership.

The court cannot produce the keys. It can still create paperwork that makes regulated intermediaries afraid to touch the coins. That is why this fight matters.

Stratum V2 1.11.0 Makes Miner Choice Less Fragile

The Stratum V2 team released version 1.11.0 with reliability work across binary handling, Stratum V1 parsing and conversion, difficulty conversion, error handling, and fuzz testing.

This is plumbing, and plumbing matters. Last month we covered DMND mining the first known mainnet block through Stratum V2 with the miner building its own block template. Miner-built templates push transaction selection away from pools and toward individual miners. That promise depends on parsers, translators, and job-declaration code that operators can trust in production.

Version 1.11.0 does not claim new pool adoption or change consensus rules. It makes miner control less fragile.

More of this, please.

vLLM Broadens the Hardware Map for Open Inference

vLLM 0.25.0 makes Model Runner V2 the default for dense models and expands work across NVIDIA, AMD, Intel, Arm, RISC-V, PowerPC, Blackwell, and Apple Silicon. It also adds broader support for speculative decoding, caching, quantization, and mTLS.

The direction is clear: open inference software is supporting a wider range of hardware.

Hardware choice is sovereignty. Wider architecture support gives builders more leverage over chip vendors, cloud providers, and geopolitical supply chains. It may also make non-NVIDIA and self-hosted deployments more practical where performance holds up.

This release removes legacy paths, so operators should stage the upgrade and test it before production. Freedom tech still needs change control.