Scarcity & Scale

Buying the AI Bottleneck

Mitchell Kochman — 2026-10-29

Written October 29, 2026. The reported acquisition talks have not resulted in an announced final agreement; subsequent developments are not reflected. Publication date: 2026-10-29. Content update date: not recorded. Views are my own; not investment advice.

Marvell is reportedly in advanced talks to acquire Lumentum, a deal that would bring together two companies that supply different parts of the infrastructure behind AI. No final agreement or purchase price has been announced.^[1]

Marvell designs chips that process and move data, including the electronics used in optical connections. Lumentum makes lasers and other optical components that carry information as light. As AI systems grow, moving data between processors becomes an important part of making the whole system work efficiently. The potential fit is a closer connection between the networking electronics and the optical components they rely on.

Marvell completed its acquisition of optical-interconnect company Celestial AI in February.^[2] Buying Lumentum would extend that push into optics. The strategic case starts with the products Marvell could build, rather than the factory it would own.

The parts have to work together. In 2025, the companies announced a demonstration combining Marvell’s electronics with Lumentum’s laser and modulator in a high-speed optical connection.^[3] They already know how to collaborate. The acquisition question is whether bringing more of that engineering and manufacturing expertise into one business would help them develop the next generation faster and make the complete connection work better.

Consider a hypothetical design choice. A change to an optical component might make it more expensive to manufacture while reducing the power consumed by the networking electronics around it. For the customer, the combined result could be worth paying for. Two independent suppliers can pursue that improvement, but they must agree on who funds the work, who bears the manufacturing risk and how each benefits. Under common ownership, management could evaluate the investment against the economics of the combined offering and assign resources accordingly.

The payoff could extend well beyond Lumentum’s own sales. Marvell can win a place in the next generation of AI systems, but those wins only turn into revenue as customers build them.^[7] If optical supply falls behind future demand, it could hold back those systems and demand for Marvell’s chips too. Owning Lumentum could help Marvell expand optical capacity alongside its silicon business, helping turn anticipated demand into shipments.

Supply agreements can accomplish a great deal of this. They can fund development, reserve capacity and establish long-term commitments. NVIDIA’s March agreement with Lumentum included a $2 billion investment, a multibillion-dollar purchase commitment and rights to future capacity. The agreement was explicitly nonexclusive.^[4] NVIDIA also announced a $2 billion investment in Marvell later that month, alongside a technology partnership.^[5] That puts NVIDIA on both sides of the proposed combination.

The additional appeal of ownership is the ability to keep making decisions as requirements change: which technology to prioritize, which manufacturing process to improve and where to spend the next dollar of expansion capital. A contract can anticipate many of those choices. An acquisition puts more of the responsibility for the choices that emerge later under the same management.

The initial market reaction brings that distinction into focus. Lumentum rose while Marvell fell in after-hours trading following the report.^[1] One possible reading is that investors see a premium for Lumentum’s shareholders while questioning the price and risks Marvell would take on. That would be consistent with a sound strategic idea that still has to prove its value to the buyer. Without transaction terms, the price moves cannot settle that question.

Marvell would also be buying the broader business: optical technology, manufacturing know-how, engineers and customer relationships. Those capabilities could support a more complete offering and give Marvell more ways to earn revenue from the infrastructure around AI processors. Whether that creates value depends on what customers actually want and what Marvell pays. The purchase price would already compensate Lumentum’s owners for the business being acquired, and Marvell would take on its operating demands and investment risks. The case needs benefits from combining the businesses beyond simply adding their revenues together.

The Greensboro factory helps explain why buying those capabilities might be attractive. In March, Lumentum announced plans to retrofit a North Carolina facility to make advanced optical components for AI data centers. Its announced target for ramping production was mid-2028, despite acquiring an existing site and experienced workforce.^[6] Building comparable capability takes time. An acquisition would give Marvell ownership of an existing optical business and a place in its expansion plans, rather than requiring it to build that expertise from scratch.

The real prize is influence over the next generation of AI infrastructure: which optical products get built, how they work with Marvell’s chips and where manufacturing capacity expands. Buying Lumentum would put more of those decisions under one roof. For Marvell’s shareholders, that is the bet: ownership makes the next generation of products better and gets them to market faster. Buying the bottleneck is the easy part. Building something more valuable around it is what makes the deal worth doing.

Sources

AI Can Find the Bugs. Who Fixes the Internet?

Mitch Kochman — 2026-10-23

By Mitch Kochman · Scarcity & Scale

Written October 23, 2026. Publication date: 2026-10-23. Content update date: not recorded. Views are my own; not investment advice.

We’ve opened Pandora’s box. The capabilities that let us build faster also let attackers find and exploit weaknesses faster. Those tools are spreading, and we cannot take them back. We now have to defend a world that wasn’t built for what they can do.

Bitcoin was always going to be a canary in the coal mine. When a security failure can lead directly to money being stolen, there is an immediate incentive to find it. The Coldcard incident this summer made that painfully clear. A firmware flaw weakened the way some wallets generated their seeds, allowing attackers to recover keys and steal funds without taking possession of the devices.

Developers then reproduced the attack with help from frontier AI models. Coinkite had already been using AI to review its code. That review missed the flaw, and its subsequent tests with frontier models missed it too. That is the asymmetry we have to understand: asking AI to find a way in is a very different task from asking it to protect against every way in. One missed weakness can be enough.

An attacker doesn't need the best AI model in the world. They need one good enough to find a weakness and turn it into a way in. Increasingly, they can download that capability and run it themselves. There is no account for a provider to suspend and no central switch that takes those copies back.

That changes the protection we get from keeping the most capable models behind restrictions. Those restrictions can buy defenders time. They cannot make the capabilities stay there.

Open-weight models are still behind the leading frontier systems overall, but they are crossing thresholds that matter. GLM-5.3 can develop working exploits in controlled tests where its predecessor barely succeeded. The frontier can remain ahead while the tools available to everyone else become dramatically more dangerous.

Attackers are already using these tools to break into financial institutions and steal data. A recent campaign against South Korean financial organizations combined an open-source agent framework with models including DeepSeek and GLM.

Now put that against the software the internet runs on.

The PHP flaws disclosed this week could allow an attacker to take control of affected servers without logging in. These weaknesses sit underneath the websites and applications people use, not just in some obscure app they could choose to avoid. The NSA recommended taking affected systems offline if they couldn't be updated immediately, until they could be fully mitigated.

Attempts to exploit PHP2Shell were detected before public disclosure, with broad attempts afterward. The full impact remains unclear, but the response window is already shrinking. Amazon is telling security teams to measure their response to critical internet-facing vulnerabilities in hours, not weeks.

AI is the only way to defend at the scale and speed we are heading toward. Human teams cannot manually match an expanding supply of machines searching for weaknesses, testing them and trying again. Defenders need AI doing that work continuously too, finding problems, developing fixes and checking whether those fixes hold.

And we are still relying on people to make that happen.

Much of our digital life rests on shared open-source software maintained by teams far smaller than the number of businesses depending on them. In PHP's case, early signs of one underlying problem were found in 2024. The work stalled over difficult trade-offs, limited specialist capacity and a decision process that couldn't resolve them quickly enough.

They had to fix a foundational system without breaking the software built on top of it. Then every organization running that software had to get the repair into use.

AI is already taking on more of that repair work. It can help write patches and test what those changes might break. But people still review the changes, approve commits and get updates deployed. The model can generate the repair faster, but getting it into use still depends on someone having the time, knowledge and authority to act.

How long is that sustainable?

More funding and better tools will help, but they do not answer the long-term question. If attackers can keep increasing the number of weaknesses they investigate, can we keep protecting the internet with a process that waits for people to approve and ship each important repair? And if we give AI more authority to make those changes itself, how do we know it isn't introducing the next failure across thousands of systems at once?

That is the reckoning ahead. We need more autonomous defense, and we need to be able to trust it with systems we cannot afford to break. Neither waiting for human teams to catch up nor handing over unrestricted control is a serious answer on its own.

The question is how quickly we can move from AI that helps people find and fix bugs to systems that can carry more of that responsibility reliably. Until then, the gap between what attackers can attempt and what people can safely repair is the one we should be watching.

Source notes

AI Agents Are Making the Case for Bitcoin

Mitch Kochman — 2026-10-22

By Mitch Kochman · Scarcity & Scale

Written October 22, 2026. Publication date: 2026-10-22. Content update date: not recorded. Views are my own; not investment advice.

An AI agent can research a business, build a product and figure out what to do next. But when that next step requires paying for something, it runs into a very human problem. Someone has to give it access to money, and the institutions holding that money get to decide how that access works.

In an October 22 post, @chamath predicted that banks would block agents to protect themselves, then need a security and governance architecture. That raises a question: who will build it, and who will control it?¹

I believe this is going to make the value of permissionless money much easier to understand. As we ask software to do more on our behalf, the ability to move value will become part of what makes that software useful. Bitcoin ($BTC) was built for a world in which two parties could transact without needing a financial institution to approve their relationship. AI is giving that idea a new audience.

The banking debate deserves some precision. On its October 13 earnings call, JPMorgan ($JPM) described blocking agents that access money movement through shared customer credentials and screen scraping. It also said it was working on doing this correctly. Handing an agent your banking password creates real risks. The bank's position leaves room for authorized agent payments.²

Even so, the underlying tension is significant. Developers can improve an agent's ability to reason and act, then find that its ability to complete a task depends on a separate institution's willingness to support it. Where that friction persists, there will be an incentive to build on payment networks that software can use directly.

Money that software can use

Imagine asking an agent to research a new business idea with a $20 budget. Along the way, it finds a useful dataset, a paid search service and a small amount of computing capacity that would improve its answer. Each purchase might cost pennies. Requiring a new account, subscription or human checkout for every service quickly consumes the time the agent was supposed to save.

An agreed price and a common payment protocol make a different experience possible. The agent can decide whether a resource is worth buying, pay within its budget and continue working. A small business selling that resource can serve a customer it has never met, without first negotiating a billing relationship.

For developers trying to connect services across companies and countries, a shared payment mechanism has considerable value.

This is how I expect useful adoption to develop. A developer encounters a problem, finds that open monetary infrastructure solves it and builds something better. The customer benefits without needing to understand everything underneath.

Where stablecoins fit

I also expect stablecoins to play a substantial role. A business that prices its services in dollars will often prefer to receive dollars, and an agent operating with a dollar budget benefits from predictable purchasing power. That is a practical advantage.

@coinbase's x402 offering gives businesses a way to integrate agent payments into an API or checkout and receive $USDC. It is a concrete example of broader crypto infrastructure making software-driven commerce easier. Businesses still have to support the payment method, but they can begin accepting payments through a common interface rather than inventing a new arrangement for each agent.³

The distinction between a programmable dollar and Bitcoin becomes important when we ask who ultimately controls the money. USDC has an issuer. @circle's terms allow it to block transfers involving certain addresses, and its reserves depend on the existing financial system. Moving dollars onto a blockchain improves how they can be used while preserving those dependencies.⁴

Bitcoin gives us a monetary asset without a company standing behind its redemption or controlling a central list of approved holders. Its network checks whether a transaction follows the rules, including whether the funds can be spent. The same rules apply whether a person constructs the transaction or software does it under that person's authority. Choosing a custodian can reintroduce a gatekeeper, even when the underlying asset is bitcoin.

That neutrality matters. As more economic activity happens through software, the question of who can interrupt it becomes more consequential. Businesses may choose programmable dollars for day-to-day expenses while valuing Bitcoin's independence for savings and settlement. These uses can develop alongside each other.

Building around the owner

The people and businesses funding agents must remain in control. An agent should have a defined budget, restrictions on where it can spend and a way for its owner to stop further activity. Payment keys belong behind controls that enforce those limits, rather than exposed to whatever instructions a model happens to encounter.

Those requirements create opportunities for builders. The useful products will make delegated spending dependable enough that someone trusts an agent with a real task and a real budget. Merchant acceptance, refunds and recourse will matter as much to many customers as the payment itself.

Traditional finance is competing for that business too. Visa ($V) is developing agent payments tied to verified funding sources and specific permissions. On October 20, Mastercard ($MA) demonstrated a system combining x402 with its own network and announced developer early access. In the pilot, an agent bought research and generated images within an approved budget, with infrastructure that included Coinbase and Base.⁵

That collaboration is revealing. Established payment companies see value in the same tools that could help developers reduce their dependence on traditional banking access. Their distribution and merchant relationships are real advantages. Open networks will have to earn adoption by making particular transactions easier, cheaper or more accessible, including through products built with those incumbents.

I think they will. The more useful agents become, the more often their developers will encounter the limits of systems built around human accounts and human workflows. Every additional service an agent can buy expands what it can accomplish. A payment network that makes those connections easier becomes more valuable to the people building on it.

We should judge that progress by working products and real commerce. More activity on a blockchain does not automatically translate into higher prices for its tokens, and Bitcoin's role will depend on the benefits it delivers.

For years, Bitcoin's independence has been easiest to appreciate when the financial system around someone fails them. AI gives us another way to understand it: as infrastructure for economic activity we are only beginning to build. I believe this is an important moment for Bitcoin because the next generation of software is making the need for open money increasingly practical. We already have the foundation. Now we have an expanding set of reasons to use it.

Sources & notes

CoreWeave Is the Headline. Hedge Fund Leverage Is the Bigger Story.

Mitchell Kochman

By Mitchell Kochman · Scarcity & Scale

Written with October 19, 2026 context. Subsequent developments are not reflected in this essay. Publication date: not recorded. Content update date: not recorded. Views are my own; not investment advice.

An AI selloff can begin with doubts about the future and accelerate because someone needs cash today.

That's what makes October's hedge-fund unwind rumors worth following. July showed how quickly a leveraged AI portfolio can run out of time and create an opportunity for whoever buys next.

The name circulating around October's AI selloff is VARA, or Value Aligned Research Advisors. The story is that another hedge fund has been caught in a forced unwind, following Situational Awareness's painful exit from much of its stock portfolio in July.

On October 15, Martin Shkreli relayed the VARA rumor, then posted a denial attributed to a person familiar a few hours later. An earlier account that named the fund subsequently called its post speculation and satire. By October 17, an account that had circulated its holdings said the liquidation story appeared wrong.

That leaves the named-fund story unresolved. The broader hedge-fund pressure is much clearer, and it is the more useful place to start.

Reuters, citing Goldman Sachs prime-brokerage notes, reported that Goldman's global fundamental long/short sample lost 2.9% in the week through Thursday, October 15. That figure does not include Friday. AI stocks had seen six consecutive sessions of hedge-fund selling, with long sales exceeding short covering.¹

The shared pressure matters: funds using different approaches can end up exposed to the same risks and trying to get smaller at the same time.

July shows how a thesis becomes a funding deadline

Situational Awareness provides a concrete example of what happens when a long-term investment thesis runs into short-term financing pressure.

Bloomberg's account of the July unwind described banks demanding more collateral as technology positions fell, followed by sales to meet margin calls. Goldman Sachs, JPMorgan and Bank of America were among its lenders. Millennium and Jane Street considered buying the portfolio before Citadel reached an agreement.²

The rate debate that week is worth putting in its full sequence. Citadel Securities strategist Frank Flight had identified July as a possible hike meeting in June and said on July 1 that markets were underpricing that risk.

Market stress was already visible on July 23, when Reuters reported a 2.15% Nasdaq decline amid AI-spending concerns and higher oil prices. On July 27, Flight made a July hike his base case. Bloomberg's July 28 coverage also reported hawkish arguments from PGIM and Wrightson ICAP.

Reuters described a split market on July 28: semiconductor stocks fell 4.5%, while the S&P 500 and Dow rose. The Fed held rates on July 29, with three members preferring a quarter-point increase.

By Wednesday morning, July 29, Citadel executives had contacted Situational Awareness, according to Reuters' subsequent reconstruction. Bloomberg reported that negotiations continued overnight and an agreement was reached before the July 30 market opening. The note came from Citadel Securities; the buyer was the separate investment firm Citadel.

Reuters described the purchase as the broker-financed portion of the public-stock portfolio, with roughly $10 billion in stocks and private investments remaining. A July 30 investor letter, reported by Reuters the following day, put the July loss at 67% and said portfolio leverage had been removed. It was still approximately 80% ahead for the year.³

The important sequence for understanding the unwind is falling positions, collateral pressure and a negotiated transfer to a buyer able to absorb the book. The reporting reviewed for this article doesn't establish a connection between the note's publication and the purchase, or that anyone deliberately triggered the unwind.

A leveraged fund does not need to abandon its view of AI to become a seller.

In a simple unhedged example, $100 of assets financed with $25 of equity and $75 of debt loses 40% of its equity when the assets fall 10%, before costs. If the lender also reduces how much it will lend against those assets, the fund needs still more cash.

CoreWeave ($CRWV), Nebius ($NBIS) and other AI infrastructure investments can remain long-term bets on demand while becoming immediate sources of liquidity for their owners.

Jordi Visser (@jvisserlabs) drew a comparison with Long-Term Capital Management in his October 17 conversation with Anthony Pompliano. The useful lesson is that a trade can move against you before it works. Your financing has to survive the wait.

What others can see before the sale

Prime brokers set and monitor collateral under their agreements with a fund. They have direct information about their own financing exposures that outsiders do not. Published accounts of Situational Awareness do not reveal the exact collateral thresholds, when each was reached, or who outside those relationships knew them.

A market does not need a complete set of loan documents to anticipate selling. Traders can see prices falling together, recognize crowded positions and reduce their own exposure. Potential buyers may wait for a larger discount. A fund trying to sell then finds less willing money on the other side.

In their paper Predatory Trading, Markus Brunnermeier and Lasse Heje Pedersen describe how anticipating another investor's forced sale can push prices beyond the initial shock. The same market can also contain ordinary risk reduction and buyers providing liquidity. A purchase after a selloff does not tell us which behavior preceded it.

In July's case, S3 Partners later reported mixed changes in short interest across the fund's disclosed long holdings, with a substantial component related to convertible hedging. That is a more complicated picture than a uniform build-up of bearish bets, although aggregate positions cannot reconstruct every individual trade.

July's reporting documents lenders protecting their exposures and buyers evaluating a distressed portfolio. More generally, research from the Bank for International Settlements describes how prime brokers tend to loosen financing in calm markets and tighten it under stress. Decisions that protect individual lenders can intensify the selling across their clients.

VARA has enough history to make the comparison catch on. Hedgeweek, relaying Bloomberg, reported that its flagship lost 44% in July, leaving it about 65% ahead for the year. Its June 30 SEC holdings filing included CoreWeave, Nebius, Bloom Energy ($BE), Lumentum ($LITE) and memory names.

But there is no comparable verified account of October collateral calls or a portfolio sale. June holdings cannot establish its positions or financing in October. The next 13F will describe September 30, also too early to settle an October liquidation claim.

Over the October 17-18 weekend, Stocktwits covered speculation about Situational Awareness's September options purchases. Older trades do not reveal its subsequent contracts, hedges or losses.

The leverage runs through both sides of the trade

Stock Analysis's closing-price histories show CoreWeave and Nebius each losing about a quarter of their value over October 9-16; Cipher ($CIFR) lost roughly a third. AP put the Nasdaq's weekly decline at about 4.5%. Those moves fit a broad retreat from the AI trade, without identifying a particular seller.

There were genuine reasons to reassess risk. The Bureau of Labor Statistics' October 14 CPI release showed September prices rising 0.6% in a month. For companies building expensive infrastructure against revenue arriving later, higher rates can change the economics. For funds borrowing to own those companies, they can also change the cost and availability of the trade.

Visser's October 18 credit discussion extends the story beyond crowded stocks. He describes high-yield bonds relative to Treasuries and convertible bonds giving back summer gains, while judging the deterioration modest compared with earlier crises. The cost of new money is worth watching alongside the pressure on existing holders.⁴

Matthews South's October 5 review described widening AI credit spreads in September and losses for convertible-arbitrage investors, alongside roughly $42 billion of third-quarter issuance. Financing pressure was visible before the VARA chatter.

AQR's explanation of convertible arbitrage describes a fund owning a bond and shorting some shares to hedge its equity exposure. Closing that trade means selling the bond and buying back the stock hedge. It does not automatically produce stock selling.

Cash needs elsewhere, new issuance and dealer hedging can create different flows. In its announcement of the October 16 convertible financing, Cipher warned that related hedging and unwinds could move its shares in either direction.

This is where a fund's funding problem can begin to affect the company it owns. Selling can make securities cheaper even if demand holds up. If weak bond prices translate into expensive new borrowing, a company still funding its buildout can suffer a real reduction in future returns.

CoreWeave and the cost of new money

The October 16 All-In discussion brings that distinction into focus. Chamath Palihapitiya (@chamath) emphasizes CoreWeave's future financing needs against fixed-price contracts. David Sacks (@DavidSacks) emphasizes strong demand, scarce capacity and the difference between corporate bonds and project financing.⁵

Both arguments matter when deciding whether a price reflects a distressed holder or a deteriorating investment.

Reuters reported a CoreWeave bond trading at roughly a 15% yield. That does not reset its existing fixed coupon or establish the rate on every loan.⁶

In March, CoreWeave announced an investment-grade facility with up to $8.5 billion of capacity, including floating-rate debt priced 2.25 percentage points above SOFR. Contract-backed financing can have different terms from unsecured corporate debt, although an earlier deal cannot tell us the current price for new money.

There are limits to that protection. CoreWeave's June-quarter SEC filing says every delayed-draw facility except DDTL 4.0 was unconditionally guaranteed by the parent; even that exception has limited guarantees. Its second-quarter results reported about $679 million of operating cash flow against $6.4 billion of property and equipment purchases. Financing access remains central.⁷

Long customer contracts make projects easier to finance, while limiting the upside from rising spot prices. Shorter contracts offer more pricing flexibility and less certainty.

CoreWeave reported new three-to-six-month contracts at higher prices in September. That is evidence of demand worth weighing against the credit concerns, rather than treating the share-price decline as a complete verdict on the business.⁸

A faster market can make the gap wider

The fund mechanism is old. The speed at which investors identify a shared trade, circulate a story and react to it is the newer concern.

In his July 19 discussion, Visser argues that agent-driven momentum, leverage and volatility-control strategies can produce what he calls "speed crashes." In the October 17 conversation, he returns to the challenge that AI agents and increasingly continuous markets pose for hedge-fund risk management.

That is a plausible way for a negative headline to travel quickly through portfolios whose future cash flows are uncertain and whose owners use leverage. The simultaneous losses across strategy groups make crowding worth investigating. They do not tell us whether AI-driven trading caused it.

AI is visibly part of the information chain. After the October 16 close, the X account SPXVIX asked Grok to verify the VARA claim, then circulated holdings attributed to Grok without a filing date. The figures matched June's disclosure. An old portfolio was being presented without the date needed to interpret it.

Those posts cannot explain trading that had already happened, but they show how the next round of market discussion can begin with incomplete context.

Where the opportunity remains

July showed that a manager can lose the ability to hold a trade while another investor is able to take it over. October's reported flows show a broader retreat, even while the story about the named fund remains disputed.

The business still has to justify the patience. For AI infrastructure, that means looking at delivery, customer payments, remaining construction costs and the terms of the next financing. A sustained financing squeeze can change what the business is worth.

For an investor who can wait, that distinction creates opportunity. A fund facing a collateral call may have to sell at a price it would never willingly accept.

If customer demand holds and the business can finance its commitments, the seller's deadline can become the buyer's advantage.

Sources & notes

How to Think Clearly About Bitcoin and Quantum Computing

Mitchell Kochman

By Mitchell Kochman · Scarcity & Scale

Written with April 2026 context; the original wording and claims are retained. Publication date has not been supplied. Publication date: not recorded. Content update date: not recorded. Views are my own; not investment advice.

For years, “quantum can break Bitcoin” has floated around as one of the internet’s favorite Bitcoin panic narratives. It is directionally true in theory and usually sloppy in practice.

Bitcoin does not rely on one giant blob of cryptography that either works or fails all at once. The part under pressure from quantum computing is mainly the signature layer: the secp256k1-based ECDSA and Schnorr signatures used to prove control of coins. A sufficiently advanced quantum computer running Shor’s algorithm could, in principle, take an exposed public key, derive the corresponding private key, and forge a valid signature. That is the real issue.

So the better framing is not “quantum breaks Bitcoin.” It is this: quantum threatens Bitcoin’s current signature assumptions, especially where public keys are already exposed. That is a much more precise statement, and once you understand it, a lot of the hysteria starts to fall away.

A lot of the latest noise traces back to Google Quantum AI’s March 31, 2026 announcement. Google published a whitepaper and blog post saying future quantum computers may be able to break the elliptic curve cryptography protecting systems like Bitcoin with fewer qubits and gates than earlier estimates suggested, and urged blockchains to begin transitioning to post-quantum cryptography.

But a resource estimate is not a countdown clock.

It means the problem may be more engineerable than people thought. It does not mean the hardware bottlenecks are solved, that a cryptographically relevant quantum computer is imminent, or that Bitcoin suddenly has an expiration date. Google made the issue harder to shrug off. It did not prove the attack is around the corner.

That, to me, is where most of the noise lives. It takes a real theoretical vulnerability, combines it with an updated estimate from a serious source, and then jumps straight to the most dramatic possible conclusion. The result is a conversation that collapses four separate questions into one blob of fear: is the threat real in theory, how close is it in practice, which coins are exposed first, and what should Bitcoin do before the threat becomes urgent?

Those are not the same question, and treating them like they are is exactly how the FUD is created.

What the threat actually is

Bitcoin’s long-term quantum problem is mostly a signature problem, not a hashing problem. The reason people focus on secp256k1 is that Shor’s algorithm targets the discrete-log assumptions behind elliptic curve cryptography. That is why so much of the current work centers on public-key exposure and signature migration rather than some sweeping idea of “rewriting Bitcoin.”

The exposure map matters.

Coins in old P2PK outputs have public keys sitting on-chain permanently. Reused addresses and previously spent outputs can also leave public keys exposed. More modern constructions can reduce exposure by keeping the public key hidden behind a hash until spend time. BIP-360’s rationale is built around reducing this kind of long-duration exposure by removing Taproot’s key-path spend from the new output type.

This is the part many people miss: not all bitcoin is equally exposed, and not all exposure is equally urgent. There is a meaningful difference between coins whose public keys are already visible forever and coins whose public keys are still hidden unless and until they are spent.

Once you understand that, the whole topic becomes less mystical and more operational.

A quick note on what a BIP is

In Bitcoin, proposed protocol changes are usually introduced through BIPs, short for Bitcoin Improvement Proposals. A BIP is not the same thing as an approved change. It is the standard format developers use to write an idea down, debate it in public, refine it, and sometimes eventually adopt it.

Publishing a BIP means a proposal has become concrete enough to discuss seriously. It does not mean Bitcoin has decided to implement it. The BIPs repository itself says publication means a proposal is in scope and meets formal criteria for the repository, not that it has community consensus or is about to be adopted.

That matters here because a lot of the quantum discussion is now happening through BIPs, which is exactly what you would want if the goal is serious thinking before a crisis instead of rushed consensus during one.

Where the serious people mostly agree

If you step back from social media and look at the people closest to Bitcoin’s cryptography, protocol design, and security, the overlap is actually pretty large.

The serious center of gravity looks something like this:

The threat is real in principle.

Panic is not warranted.

Preparation is warranted.

That is a much more useful debate than the usual “Bitcoin is cooked” versus “nothing to see here” circus.

Even the current BIP discussion reflects that posture. BIP-361 is explicitly framed as following the implementation of some post-quantum output type first, and as a pre-announced sunset of legacy ECDSA and Schnorr signatures rather than an immediate activation push.

That is what serious systems are supposed to do: think through ugly problems before they become emergencies.

Where the real disagreement is

The actual disagreement is about timeline, urgency, and tradeoffs.

One camp thinks the market is factoring in way too much confidence about how quickly quantum hardware will scale. Their view is that the leap from “this is possible on paper” to “this is dangerous in practice” remains enormous, and that engineering reality is still being badly underestimated.

Another camp thinks that even a modest probability is enough to justify serious work now, because migrations in Bitcoin happen over years, not weekends, and credible preparation matters in its own right.

That disagreement is more substantive than most people realize. It is not just optimism versus pessimism. It is a clash between two intuitions that can both be reasonable at the same time: first, hardware timelines are highly uncertain and current panic may be overfitting to resource-estimate headlines; second, waiting for certainty in a system as large and inertial as Bitcoin is a bad strategy even if the probability is low.

BIP-360 in plain English

BIP-360 is best understood as a “stop making the problem worse” proposal.

It introduces a new output type, Pay-to-Merkle-Root, designed so Bitcoin can preserve useful script-tree functionality while removing Taproot’s quantum-vulnerable key path. The BIP says its quantum resistance comes from removing the P2TR key-path spend, and that users can migrate script trees from P2TR outputs to P2MR outputs for protection against long-exposure quantum attacks.

In plain English, it is a way to reduce future long-duration public-key exposure in new outputs.

What BIP-360 solves is future exposure. It gives users a cleaner path to avoid creating more long-lived public-key exposure going forward.

What it does not solve is the much touchier legacy problem, which includes old P2PK outputs, among them many of the Satoshi-era coins whose public keys are already exposed on-chain.

It is scaffolding, not salvation.

That modesty is a feature, not a bug. BIP-360 tries to make forward progress on one narrow but important part of the problem without forcing immediate consensus on every harder question downstream.

BIP-361 in plain English

BIP-361 is where this conversation gets a lot more charged.

The draft proposal is called Post Quantum Migration and Legacy Signature Sunset, and the core idea is simple: if Bitcoin eventually has a credible post-quantum destination, the network could pre-announce a phased sunset of legacy ECDSA and Schnorr signature paths. The draft says Phase A would disallow sending funds to quantum-vulnerable addresses, while Phase B would later make ECDSA and Schnorr spends invalid for those legacy UTXOs.

In plain English, it is an attempt to avoid waiting until an actual emergency to decide what to do with coins that remain vulnerable under the old signature model.

A huge part of why BIP-361 touched such a nerve is that this is not really about some random future wallet problem. It includes the most iconic dormant coins in Bitcoin history. Google’s whitepaper explicitly highlights old P2PK outputs as a quantum-vulnerable bucket because their public keys are already on-chain, and outside analysis of that paper notes this includes more than 1.7 million BTC from the Satoshi era.

Inside that bucket sits the number everyone immediately thinks about: the roughly 1.1 million BTC commonly attributed to Satoshi. That 1.1 million figure is an estimate, not something proven by Google’s paper itself, but it is the widely cited number in mainstream coverage and market reporting.

That is a big part of why BIP-361 is even being discussed.

BIP-360 is a forward-looking step. It helps Bitcoin stop creating more long-duration public-key exposure in new outputs. But it cannot do anything for coins whose public keys were revealed 15 years ago and are now exposed forever. BIP-361 is aimed at that uglier legacy problem: what, if anything, should Bitcoin do about old exposed coins before a real quantum threat turns them into a standing bounty?

And once you say that out loud, the debate stops being academic. Now you are no longer just talking about cryptography or migration paths. You are talking about whether Bitcoin should ever take away a participant’s ability to move and use coins they still control under the existing rules. That is a huge deal. For a network that prides itself on immutability, censorship resistance, and the idea that valid keys give you valid access to your property, stranding vulnerable legacy coins would be a line Bitcoin has never really crossed before. Even if the argument is made in the name of protecting the broader system, it is easy to see why many people view that as an extraordinary and uncomfortable precedent.

That is why BIP-361 hit such a nerve. It is not just a cryptography debate. It is a Bitcoin values debate.

The best criticism of BIP-361 is not outrage-posting. It is that Bitcoin should be extremely careful about any rule set that effectively strands coins based on address type or cryptographic vulnerability, because that moves the system into uncomfortable philosophical territory.

The best defense is that existential threats do not care about clean ideology, and that the market may prefer ugly preparation to catastrophic theft.

That tension is the real heart of the debate.

Why Google’s paper poured gasoline on this

The Google paper mattered because it narrowed the distance between “theoretical vulnerability” and “conceivable attack plan.” Google’s blog explicitly says future quantum computers may break the elliptic curve cryptography protecting cryptocurrency with fewer qubits and gates than previously realized, and recommends transitioning blockchains to post-quantum cryptography before that becomes possible.

When a company like Google says the resource estimates may be dramatically lower than many people thought, it is not surprising that markets and media start to extrapolate aggressively.

But the overreaction comes from treating “lower than before” as “close enough to price like certainty.”

That does not follow.

Resource estimates still sit on top of assumptions about error correction, architecture, connectivity, clock speed, and the sheer difficulty of scaling fault-tolerant systems. The Google paper made it harder to shrug off the issue. It did not prove that a practical attack is around the corner.

This is why the best response is neither dismissal nor panic. The right takeaway is that the threat has become harder to hand-wave away, while the timeline remains too uncertain for apocalyptic certainty.

Both halves matter.

Bitcoin is not asleep at the wheel

One of the laziest versions of the current quantum hysteria is the idea that Bitcoin is doing nothing.

That is just false.

The BIP-360 workstream exists. The BIP-361 draft now exists. There are active proposals and public design discussions around migration paths and post-quantum approaches. The existence of both proposals in the BIPs repository is direct evidence that Bitcoin developers are already debating the tradeoffs in public.

That does not mean Bitcoin has solved the problem. It means the system is doing what resilient systems are supposed to do: debating ugly edge cases before they become emergencies.

If anything, the current conversation should make people more confident that the hard parts are being argued over now, while there is still room for caution, iteration, and disagreement.

The deepest issue is not technical

The hardest part of this conversation is not “can Shor’s algorithm break elliptic curve cryptography?”

At the level that matters here, the answer is yes in principle.

The harder question is what Bitcoin should be willing to do in response, and when.

Should users be left to migrate voluntarily forever?

Should the protocol ever pressure them?

Should it ever strand old coins if the alternative is large-scale theft?

Those are not engineering questions alone. They are questions about what kind of system Bitcoin wants to be under stress.

That is why this debate gets emotional so quickly. People are not just arguing about qubits and signatures. They are arguing about property rights, social contract, precedent, and whether preserving Bitcoin’s core ethos sometimes conflicts with protecting Bitcoin’s broader economic integrity.

The right takeaway

Bitcoin’s quantum risk is real in principle. The confusion is in the framing.

The lazy version of the story says quantum can break Bitcoin, therefore Bitcoin is suddenly on a short fuse.

The more honest version is this:

A sufficiently powerful quantum computer would threaten Bitcoin’s current signature assumptions, especially where public keys are already exposed. Google’s recent paper made that threat feel more concrete. The hardware path is still uncertain. And Bitcoin is already exploring ways to reduce exposure and prepare for a harder migration if it ever becomes necessary.

The smartest voices in Bitcoin are not really debating whether the abstract vulnerability exists. They are debating timeline, exposure, migration, and governance.

How soon could this matter?

Which coins are first in line?

What should Bitcoin change now?

What should it refuse to change even under pressure?

That is the real conversation, and it is far more useful than the usual doom loop.

If I had to boil it down to one line, it would be this:

The quantum threat is worth preparing for, but not worth panicking.

Bitcoin is not asleep. It is doing what serious systems do when the stakes are high and the timeline is uncertain: thinking early, arguing hard, and trying not to confuse possibility with inevitability.

What If Bitcoin Is Driven by Real Yields, Not the 4-Year Cycle?

Mitchell Kochman

By Mitchell Kochman · Scarcity & Scale

Bitcoin was built for a world where cash stops working in real terms.

For years, Bitcoin holders have relied on a familiar script.

Every four years, the halving cuts new supply in half. Price responds. Momentum builds. New capital enters the market. A bull cycle follows. Then comes excess, leverage, the blow-off top, and the inevitable reset. Repeat.

It is a clean story. It is intuitive. It is easy to remember.

It also may never have been the real driver.

That does not mean halvings do not matter. They do. Bitcoin’s fixed supply and declining issuance schedule are foundational to the asset. They are part of what makes Bitcoin different from fiat currencies, from commodities with elastic production, and from most other financial assets in the modern system.

But there is a difference between something being foundational and something being the primary driver of price. I’m beginning to believe that the macro environment has been driving us the whole time, with the 4 year cycle being more of a coincidence. Bitcoin has often done best when the real return on cash is deteriorating, the policy backdrop is becoming less restrictive, and liquidity conditions are improving. In that framework, the halving still matters, but more as an accelerant than the engine itself.

I found Jordi Visser’s recent article “How AI, Inflation, and Scarcity Are Driving Bitcoin Into Its Strongest Regime” fascinating as it compares bitcoin performance to the yield of 3 month treasuries. His core claim is straightforward: Bitcoin’s strongest historical returns came when the Fed was on hold or easing and real 3-month rates were at or below zero.¹

The reason he focuses on 3-month yields is that they are one of the clearest measures of what investors can earn in short-term cash-like instruments. In practical terms, they represent the return available in very short-duration government paper — the kind of place investors hide when they want safety, liquidity, and yield. When those short-term yields are above inflation, cash preserves purchasing power and becomes a real competitor to Bitcoin. When inflation rises above those yields, cash may still look safe nominally, but it starts losing value in real terms. That is the shift Jordi is highlighting.

In his framework, the key issue is not just rates in isolation, but real short-term rates — in other words, short-term yields relative to inflation. His argument is that Bitcoin tends to struggle when investors are being paid a compelling real return to sit in cash, because the hurdle rate for owning a volatile, non-yielding asset is high. But when inflation overtakes those short-term yields, that equation changes quickly. Cash stops functioning as a true store of value in real terms, and capital has more reason to look for scarce alternatives. That is why Jordi identifies the strongest regime as the one where CPI is above 3-month bills and the Fed is on hold or easing.

What makes that framing compelling is that it gives a macro explanation for Bitcoin’s major moves. Instead of saying Bitcoin rises because of a fixed four-year calendar, Jordi is saying Bitcoin rises most when the real return on cash breaks down and policy is no longer tightening against risk assets.

It suggests Bitcoin does not simply respond to supply mechanics. It responds to whether investors are being paid, in real terms, to sit in cash.

Jordi’s chart below captures this point: Bitcoin tends to outperform when investors are no longer earning a compelling real return in short-duration instruments. That is exactly the kind of backdrop Bitcoin was built for — a world where nominal safety starts to disappoint in real terms, and scarce monetary assets become more attractive.

Open chart at full size

Figure 1. Jordi Visser’s regime matrix. Bitcoin’s strongest historical returns came when the Fed was on hold or easing and real 3-month rates were at or below zero. The strongest box in the matrix shows +247% annualized returns.

I wanted to dive more into this matrix and Bitcoin’s relationship with real yields. I found the more actionable signal may be that Bitcoin improves when the real return on cash is moving lower and becoming less restrictive rather than the real yields themselves.

Open chart at full size

Figure 2. Bitcoin vs. a real 3M T-bill proxy since 2017. The cleaner read is not simply negative versus positive real rates, but whether the real return on cash is becoming less attractive.

A move from strongly positive real yields to barely positive real yields can still matter. Even if cash remains nominally attractive, the opportunity cost of owning Bitcoin is no longer rising. The headwind is fading. Markets often begin repricing that change before a simple zero line is cleanly crossed.

That is why the Bitcoin-versus-real-3M chart matters so much. It does not just separate negative from positive real 3-month rates. It shows the broader arc: when the real return on cash becomes more punitive for Bitcoin, performance tends to struggle. When that pressure eases, Bitcoin tends to improve. The cleaner read is not simply negative versus positive. It is whether the real return on cash is becoming less attractive on a trend basis.

The halving narrative became canonical because the timing often looked clean enough to feel self-evident: halving, rally, peak, reset. But correlation is not causation. It is entirely possible that the halving cycle worked as a story because it repeatedly overlapped with improving liquidity conditions and a more supportive real-rate backdrop.

In that framework, the halving still matters. But it matters more as an accelerant than as the true engine.

Figure 2. Bitcoin vs. a real 3M T-bill proxy since 2017. The cleaner read is not simply negative versus positive real rates, but whether the real return on cash is becoming less attractive.

Why this may matter now

This is where the current setup becomes especially important.

The case is not that real short-term rates are already deeply negative. The case is that they may be moving toward zero quickly enough for markets to start repricing the regime before the line is cleanly crossed.

That matters because inflation is no longer a theoretical risk. March CPI rose 3.3% year over year, up sharply from 2.4% in February, while the energy index rose 12.5% over the same period. Oil is a big part of that story. After beginning the year near $61 per barrel, Brent crude finished the first quarter around $118 as conflict in the Middle East and disruption around the Strait of Hormuz pushed energy prices sharply higher.

That creates a meaningful macro tension. If energy is pushing inflation higher while the labor market softens, then the real return on cash can erode quickly even before nominal front-end rates move much. Cash does not have to become obviously unworkable overnight. It just has to become less compelling.

The Fed may also have less room to respond than in a typical inflation scare. The March FOMC minutes explicitly noted that labor-market conditions appeared vulnerable to adverse shocks and that a further drop in labor demand could push unemployment higher in a low-hiring environment.

That is where AI enters the picture. Not as a neat forecast, but as a plausible amplifier. If AI-driven productivity gains are accompanied by faster displacement in software, services, and knowledge work, then the Fed could find itself facing a difficult mix: inflation pressure through energy and physical scarcity on one side, and weaker hiring and rising unemployment on the other.

If that is the backdrop now emerging, then Bitcoin does not need perfect conditions to become more attractive. It simply needs the real return on cash to stop looking safe enough. That is the kind of inflection markets tend to price before the data look fully obvious.

This is the kind of environment Bitcoin was built for

This may be the most important point.

Bitcoin was not built for a world where cash safely preserves purchasing power, trust in policy remains high, and the monetary system feels stable and sufficient.

Bitcoin was built for the opposite.

It was built for an environment where the return on fiat money disappoints in real terms. Where monetary discipline weakens. Where holding cash feels safe nominally but destructive in purchasing power. Where investors begin to realize that scarcity matters more than promises.

That is Bitcoin’s natural habitat.

Bitcoin is fixed supply in a world of expanding balance sheets.

It is monetary certainty in a system built on discretion.

It is an asset that asks investors to step outside the logic of managed currency debasement and political convenience.

That is why the real-rate discussion matters so much. When real cash yields are rising, the system is effectively paying investors to remain inside the fiat architecture. When those real yields flatten or deteriorate, that payment weakens. The appeal of an alternative monetary asset rises.

That is when Bitcoin begins to look less like a speculative side bet and more like what it was always meant to be: scarce, apolitical money for a world where the existing system no longer feels trustworthy in real terms.

A second tell: Bitcoin may be separating from software

There is a second layer to all of this, and it may be arriving at exactly the right time.

For much of the past few years, Bitcoin often traded like long-duration tech beta. That reinforced the idea that it was just another code-based risk asset.

But if Bitcoin is beginning to hold up better than software while the real return on cash deteriorates, that is a different signal.

If that relationship is changing, Bitcoin may be shifting away from the speculative tech bucket and back toward its original monetary identity: scarce digital property in a world where the return on fiat is weakening.

That would be a meaningful shift.

Open chart at full size

Figure 3. BTC vs. IGV normalized performance, March 12-April 10, 2026. Over this recent stretch, Bitcoin held up materially better than software, hinting that it may be separating from the long-duration tech bucket.

The takeaway

The 4-year cycle was always the easy story.

The real story may have been cash getting less attractive.

The halving mattered. But macro likely mattered more. And within macro, the direction of real short-term yields may be one of the clearest signals to watch.

If that direction is now turning lower, Bitcoin may be moving back into the kind of environment it was built for.

And if that is true, the next major move in Bitcoin may have far less to do with the calendar than people are ready for.

Sources & notes

“How AI, Inflation, and Scarcity Are Driving Bitcoin Into Its Strongest Regime”. ↩