Fox's $22 billion acquisition of Roku reprices the streaming stack as the Bank of Japan lifts rates to 1.0% this morning, its highest since 1995, with Governor Ueda absent for the first time in BOJ history. Warsh's first FOMC convenes today with a hold locked in, while the Iran nuclear deal awaits its Friday signature in Switzerland. Goldman Sachs projects $7.6 trillion in cumulative AI capital spending through 2031.
The Bank of Japan's hike is no longer a forecast. The decision has landed, and Markets & Macro below carries the vote, the dissent, and what the forward guidance means for global capital. What is settled is the move; what is still open is the path.
Asia took it unevenly. The Nikkei briefly topped 70,000 for the first time before trimming to +0.6% (69,713) as the yen firmed and JGB yields rose, the Kospi pushed further into record territory (+2.1%), and the Hang Seng slipped 1.3%. The split tape is the carry trade repricing in real time, the thread Markets & Macro picks up below.
Crypto data provided by CoinGecko
The Bank of Japan delivered its widely expected hike to 1.0% this morning in a 7-1 vote, with Toichiro Asada the lone dissent for a hold, lifting the policy rate to its highest since 1995. Governor Ueda, hospitalized with a hepatic cyst infection, submitted his views in writing but did not vote, the first BOJ meeting ever held without its governor. With the hike confirmed, the live question is no longer whether the BOJ moves but whether this is the peak or the midpoint. Ueda's written statement carried less signaling power than a press conference, which left the market to parse the board's official statement for normalization clues without the usual Q&A filter, and the initial verdict offered little dovish offset. If the path from here signals further tightening, the implications shift from tactical to structural: Japan running positive real rates for the first time in decades would pull trillions in capital home from foreign bonds and equities, tightening global conditions independently of anything the Fed does.
Warsh chairs his first FOMC meeting starting today, with a hold near-certain and the real variables being the dot plot and the new chair's press-conference language. The market is positioned long dollars at $27.8 billion in spec longs, the highest since February 2025, while the DXY sits below 100. That divergence tells you traders expect a hawkish hold to validate their bets while the price says the BOJ-driven yen rally is the dominant force regardless. May's CPI at 4.2% and PPI at 6.5% give Warsh ammunition to hold firm, but the question markets need answered is whether he frames the path forward as "inflation first" or "balanced risks." If the statement leans hawkish, the crowded dollar long holds but the move is already priced. If Warsh acknowledges growth risk alongside inflation, the long gets squeezed into a yen rally and the dollar breaks below its 2025 range.
Fox Corp announced its acquisition of Roku, offering $160 per share ($96 cash and $64 in Fox stock) and creating a combined entity where Fox controls roughly 73% of the business. The headline reads like a media company buying a streaming platform. The structural read is different: Fox paid for the remote control. Roku's 100 million active accounts, its SmartTV operating system embedded in roughly a third of US television sets, and its first-party data stack represent the viewing surface. Fox is betting that owning the surface is worth more than competing for the content. At roughly $220 per household for a permanent screen position, versus Disney's roughly $53 per subscriber per year in content costs with no retention guarantee, Fox is buying distribution and plugging its existing properties (Tubi, Fox Sports, Fox News) into it, skipping the content-spending treadmill. This follows the same logic as S&P Global absorbing IHS Markit: whoever controls the interface where participants meet captures the margin, even without producing the underlying product.
Morpho, a DeFi lending protocol that optimizes capital allocation across multiple lending platforms, raised $175 million at a $2 billion valuation in a round co-led by a16z and Paradigm. The raise values a protocol that acts as a yield-routing layer, sitting above platforms like Aave and Compound to allocate deposits where risk-adjusted returns are highest. That valuation marks a shift in how DeFi infrastructure gets priced: not on total value locked, which for Morpho is modest, but on the routing and optimization layer that makes existing protocols more capital-efficient. The parallel in traditional finance is the prime brokerage layer that emerged above exchanges in the 1990s, where the routing infrastructure became more valuable than any single venue.
Aave's governance forum is debating a new risk-management framework proposed by LlamaRisk, triggered by a nine-figure exploit on KelpDAO's restaked-ETH bridge that exposed cross-protocol contagion risk across DeFi lending. The proposal would formalize risk assessment procedures modeled on Basel-style capital requirements, including mandatory reserve buffers and automated risk scoring for collateral types. The structural shift is worth noting: DeFi protocols are migrating from reactive incident response (pause the pool after the exploit) to proactive risk architecture (price the contagion before it arrives). Whether Aave's governance can enforce risk standards across a permissionless system is the open question, but the attempt itself signals the sector is building institutional plumbing that could survive regulatory scrutiny.
Goldman Sachs Research's projection of cumulative AI-related capital spending through 2031 covers hyperscaler data centers, semiconductor fabrication, power-grid upgrades, and cooling infrastructure, and the number is more useful as a denominator than a headline. It implies the AI buildout is roughly 2.5 times the capital deployed during the US shale boom, with about 40% expected to land in the first three years. The practical question is survival through the first revenue miss. AWS, Azure, and Google Cloud collectively earned roughly $240 billion in cloud revenue in 2025, so the capex projection implies AI workloads need to more than triple from current levels. That demand curve has not been demonstrated at scale. If the spending keeps accelerating while cloud revenue growth decelerates, the buildout starts resembling fiber-optic overbuild in 2000 rather than smartphone infrastructure in 2008.
NVIDIA and SK hynix signed a multi-year technology partnership to co-develop next-generation memory architectures for AI chips, starting with HBM4 specifications. AI accelerators are increasingly bottlenecked by memory bandwidth rather than compute, and NVIDIA's latest chips devote more die area to memory interfaces than to processing cores. The partnership locks in SK hynix, which holds roughly 50% of the high-bandwidth memory market, as a strategic co-development partner rather than a commodity supplier. Samsung, the number-two HBM producer, has struggled with HBM3E yield rates, which makes SK hynix's position more durable. This is a supply-chain architecture choice: NVIDIA is vertically integrating its memory relationship the way Apple integrated its silicon relationship with TSMC, converting a component vendor into a co-designer with shared roadmap visibility.
The Iran nuclear deal moved from Sunday's contested announcement toward a signed agreement, with Pakistan confirming the ceremony for Friday and Trump authorizing removal of the US naval blockade after the signature. The contradictions that surfaced Sunday, with Iranian parliamentarians disputing whether Khamenei's four red lines were met, have not been publicly resolved. The gap between announcement and signature is historically where Middle East frameworks collapse. But the market has already moved: oil settled below $81 on Monday, stripping out supply-disruption premium that would reprice sharply if the deal fails before Friday. The structural question is no longer whether the framework is real but whether five days is enough time for both sides to paper over the contradictions their own institutions surfaced publicly. If the signing holds, Hormuz reopens and the energy market enters a new regime. If it collapses, the reversion is sharp because the peace premium is already priced.
Israeli jets struck targets in Beirut's southern suburbs overnight, killing at least three people and prompting forced-displacement orders for 29 towns in southern Lebanon. The strikes extend an aerial campaign into Hezbollah-controlled neighborhoods that any ceasefire framework would need to cover and test the durability of the broader Iran agreement. The deal presumes a regional de-escalation, but Israel's operations in Lebanon follow their own operational logic. Whether Jerusalem accepts constraints on its Lebanon campaign as a condition of the Iran framework, or continues the campaign on a separate track, determines whether the "regional peace" framing survives contact with the actual region.
The G7 summit opens in Evian today with the Iran deal, Ukrainian reconstruction, and China's industrial overcapacity dominating the agenda. Sunday's EU-China trade-and-tech consultations produced no agreement on EV tariffs, and the European Commission is preparing a separate safeguard investigation into Chinese steel. The summit's structural test is whether the seven can coordinate a unified response to Beijing's overcapacity exports. A joint statement with specific enforcement mechanisms would signal coordinated action. Absent that, the summit produces communiques and the trade response remains fragmented.
Materials scientists at MIT demonstrated a concrete formulation that mineralizes atmospheric CO2 into its pore structure over time, gaining roughly 12% compressive strength in its first year while sequestering approximately 0.5 kilograms of carbon dioxide per cubic meter of material. Conventional Portland cement production accounts for roughly 8% of global CO2 emissions, and the material degrades under carbonation. The new formulation inverts this relationship: the same atmospheric CO2 that weakens conventional concrete actively strengthens this variant by converting gaseous carbon dioxide into solid calcium carbonate within the cement matrix. If the chemistry scales to commercial production, every building, bridge, and road surface becomes a slow-acting carbon sink rather than an emissions liability. The question is whether the 12% strength gain holds at industrial pour volumes and outdoor curing conditions, which introduce variables that laboratory samples do not encounter.
Marine archaeologists discovered the wreck of the Hofuku Maru off the Philippines, a Japanese freighter sunk by US bombers on September 21, 1944, while carrying 1,047 Allied prisoners of war who were killed in the attack. The bombers did not know prisoners were aboard because Japan systematically refused to mark POW transport ships under international convention. Known as "hellships," these unmarked prisoner transports were among the most lethal aspects of the Pacific War. The wreck was located using multibeam sonar at roughly 150 meters depth, providing physical evidence for an event previously documented only through survivor accounts from nearby ships.
Entomologists at Cornell documented approximately 5.5 million ground-nesting bees of the genus Andrena in a single cemetery in Ithaca, New York, the largest known aggregation of solitary ground-nesting bees in North America. The bees, each constructing individual nests in the soil, collectively occupied about 2.5 acres of cemetery lawn that had gone decades without pesticide treatment. The discovery challenges the assumption that solitary bees are functionally antisocial: they aggregate in enormous numbers when soil conditions, floral resources, and the absence of chemical treatment align. Modern landscaping practices, particularly herbicide use and frequent mowing, destroy exactly the conditions these aggregations require.
Researchers built an artificial photosynthesis system that self-regulates its light absorption the way biological leaves do, protecting itself from photodamage during intense sunlight and resuming full operation when conditions normalize. The system pairs perovskite light absorbers with molecular catalysts to convert CO2 and water into fuel at rates approaching natural photosynthesis. The self-repair mechanism is the breakthrough: previous artificial photosynthesis systems degraded within hours under real-world light conditions, lasting nowhere near the months needed for practical deployment. A system that survives its own operating environment is the difference between a laboratory demonstration and deployable energy technology.
Copper exchange inventories have drawn down to levels not seen since 2005, while every data center under construction requires roughly four times more copper wiring than a conventional commercial building.
Global copper inventories on the LME and COMEX combined sit near 300,000 metric tons, the lowest in two decades, representing roughly 2.5 weeks of global consumption. The structural deficit has three converging drivers. First, no major greenfield copper mine reaches full production before 2028, with Kamoa-Kakula Phase 3 in the DRC and QB2 in Chile the only material near-term additions to supply. Second, the pipeline of planned data center capacity exceeds 35 GW globally through 2028, and each gigawatt requires approximately 20,000 metric tons of copper for wiring, bus bars, transformers, and cooling infrastructure. Third, the energy transition, covering EV production, grid upgrades, and renewable installations, consumes copper at rates that compete directly with the data center buildout for the same constrained supply. If copper inventories breach 200,000 metric tons while data center construction continues accelerating, expect copper to break above $12,000 per metric ton from the current level near $9,500, which flows through to higher costs for every piece of electrical infrastructure from server racks to residential wiring. Watch: LME and COMEX copper inventory reports monthly, and any data center construction delays linked to copper availability or electrical infrastructure bottlenecks.
The first wave of post-COVID office lease expirations hit in 2024 and 2025, and the second wave begins this quarter as corporate leases signed during the 2012 to 2016 expansion reach their 10 to 15 year renewal dates.
The first wave produced downsizing: tenants renewed at 20 to 30% smaller footprints or shifted to flexible coworking space. The second wave may produce outright vacancies, because many of these leases were signed by companies that have since adopted permanent hybrid or remote work policies. Class B and C office vacancy rates in secondary metros like Austin, Phoenix, Nashville, and Charlotte already sit near 22%, up from 14% pre-COVID. The structural problem is not demand destruction but demand migration: the need moved to a different building type (flex and coworking) and geography (suburban), and the existing Class B office stock in urban cores cannot follow it. If vacancy rates in these secondary metros exceed 25% by Q4 2026, the CRE CLO market faces its first ratings-downgrade cycle since 2010, which transmits through regional bank balance sheets still carrying office loans marked at pre-COVID valuations. Watch: Cushman and Wakefield and JLL quarterly vacancy reports for secondary metros, and Moody's and S&P commercial mortgage-backed securities watchlist additions through Q3 2026.
Surface Capture: in any layered media stack, value eventually migrates from the content layer to the distribution layer, and whoever controls the viewing surface captures the margin. Fox's Roku acquisition is a bet that the screen is worth more than what plays on it.
Roku does not make hit shows. The Roku Channel is a free, ad-supported aggregator that nobody seeks out by name. What Roku controls is the viewing surface: 100 million active accounts, a SmartTV operating system embedded in roughly a third of US television sets, and a first-party data stack that knows what those households watch, when they watch, for how long, and which ads they respond to.
The acquisition is not a bet on content. It is a bet that the household relationship is the scarce asset in streaming. Fox is buying the distribution surface and plugging its existing properties (Tubi, Fox Sports, Fox News) into it, bypassing the content-spending treadmill that has cost Disney, Warner, and Paramount tens of billions with uncertain retention. This is surface capture: a repricing of what "television" is actually worth, where the interface layer absorbs the margin that the production layer assumed was permanently its own.
The precedent is not other media M&A. It is S&P Global's $44 billion acquisition of IHS Markit: whoever controls the interface where participants meet captures the margin, regardless of who produces the underlying product.
Where this breaks. Fox has no history of operating a technology platform. Roku's SmartTV OS requires continuous engineering investment, and Fox has no engineering culture to speak of. If Fox fails to retain Roku's product and engineering talent through the integration, it bought a depreciating asset at peak price. More fundamentally, the viewing surface may already be commoditized. Google TV, Amazon Fire TV, Samsung Tizen, and LG webOS collectively control more distribution than Roku. If the screen is a commodity and content is what differentiates, Fox paid a massive premium for a commodity position. Netflix runs on every screen and still commands premium pricing, which suggests content, not surface, drives the willingness to pay. Test: Fox Q1 2027 combined-entity ad revenue per household versus Roku's standalone ARPU of $43.63 in Q4 2025. If flat or declining, the surface was never the margin and the deal was a defensive overpay.
"In strategy it is important to see distant things as if they were close and to take a distanced view of close things."
— Miyamoto Musashi, The Book of Five Rings
The weekly preview is already open. You are mapping three time zones, two central bank decisions, and a deal that could rewrite oil markets by Friday. The temptation is to go granular immediately: start modeling BOJ scenarios, calculate the oil-price asymmetry, draft three versions of the client memo.
But the person who navigates dense weeks well is not the one who analyzed each event at the highest resolution. It is the person who stood back far enough to see which events were actually connected and which ones could be ignored entirely. The close thing screaming for attention this morning might be noise by Wednesday. The distant thing you have not examined yet might be the only one that changes a decision.
Here is the uncomfortable version: the report that needs your attention is not the one with the deadline. It is the quiet one where your conclusion would change someone else's next move. You keep putting it off because it requires thinking you have been avoiding.
Today's practice: before you touch the most urgent item on your list, spend five minutes on the one you have been deferring. Write the first sentence of the conclusion, even if you are not sure it is right. Starting with the distant thing, not the close thing, is the discipline Musashi is describing.
In 1971, economist George Stigler published a paper that inverted a century of assumptions about regulation. The standard view held that regulatory agencies existed to protect the public from industry. Stigler showed that, over time, industries reliably captured the agencies designed to regulate them and converted regulatory power into competitive advantage. The protector becomes the weapon.
The mechanism works through an asymmetry of attention. The public cares about airline safety diffusely and intermittently. The airlines care about FAA rulemaking intensely and continuously. The people who show up to every hearing, who provide technical expertise to understaffed agencies, who hire former regulators at triple their government salary, and who fund the political campaigns of oversight committee members are not the public. They are the regulated. Over a decade, the agency's staff, institutional knowledge, career incentives, and default assumptions tilt toward the industry's perspective, not through corruption but through proximity.
The Interstate Commerce Commission, created in 1887 to protect farmers from railroad monopoly pricing, spent its final decades protecting railroads from trucking competition. The Civil Aeronautics Board, created to ensure safe and affordable air travel, spent decades preventing price competition and blocking new airline entrants. Both agencies were eventually abolished, and in both cases the industries they were ostensibly regulating fought hardest to preserve them.
The counterintuitive conclusion is that the more powerful you make a regulator, the more valuable it becomes to capture, which means strengthening an institution's authority can make the problem it was built to solve worse. The solution is not deregulation but structural rotation: term limits for commissioners, mandatory cooling-off periods before industry employment, and periodic sunset clauses that force re-justification of the agency's existence. The goal is to make capture expensive, not to eliminate regulation.
The decision tool: when evaluating any institution that governs access, quality, or competition, ask who benefits most from its continued existence. If the answer is the entities being regulated, the guardian has already become the gatekeeper.
In 1896, American psychologist James Mark Baldwin proposed something that seemed to bridge the gap between Lamarck and Darwin without breaking either theory: a mechanism by which learned behaviors could influence evolution without any direct transmission of acquired traits.
The logic is elegantly simple. An organism that can learn to cope with a new environmental challenge, say a novel predator or an unfamiliar food source, survives long enough to reproduce, even if it was not born with the optimal genetic response to that challenge. This learning ability acts as a bridge. It keeps the lineage alive while natural selection slowly accumulates the genetic mutations that would make the learned behavior innate. Over many generations, what the ancestor had to learn, the descendant does instinctively. The capacity to learn created the conditions for evolution to hard-code the lesson.
The idea was largely ignored for nearly a century until computational biologists Geoffrey Hinton and Steven Nowlan demonstrated it in 1987 using neural network simulations. Their model showed that populations with learning ability evolved optimal solutions orders of magnitude faster than populations without it. Learning did not replace evolution. It guided it, by keeping learners alive in the neighborhood of the right solution while selection closed the remaining gap.
The implication extends well beyond biology. In any system where agents can learn, whether markets, companies, or teams, the capacity to learn is more valuable than any specific capability, because it buys time. A company with a mediocre product but fast learning loops will outperform a company with a superior product but rigid processes, not because the mediocre product wins, but because the learning infrastructure generates better products faster than the rigid competitor can respond. The fastest learners create selection pressure that rewards learning itself, which means the meta-skill compounds in a way the object-level skill cannot.
Before investing in a specific capability, ask whether you should invest in the learning infrastructure that would make every future capability cheaper to acquire.