April 08, 2026
msprvs1=20557a5C2VatU=bounces-280172-1781@bounce.news.bloomberg.com (msprvs1=20557a5C2VatU=bounces-280172-1781@bounce.news.bloomberg.com) · rss · 13 mins
When Tapestry (Coach, Kate Spade) tried to acquire Capri Holdings (Michael Kors, Versace), both companies told the public the deal was “pro-competitive” and the FTC “fundamentally misunderstands the marketplace.” The FTC sued, won, and the merger died. Capri shareholders then sued for securities fraud — and a federal judge dismissed the case, ruling “not everything is securities fraud.” Expressing a genuinely held position that turned out to be legally wrong does not constitute fraud; contrast with Elon Musk’s tweets falsely suggesting he’d exit the Twitter deal, which courts found were actionable because his position was more clearly pretextual.
Bill Ackman is framing a deal to buy ~6–7% of Universal Music Group as a “100% acquisition” at a €55 billion valuation — a number that requires some creative arithmetic. Pershing Square SPARC Holdings would contribute ~€2.5 billion in equity (€1.4B from Ackman’s funds, €1.1B from SPARC investors who currently hold “roughly zero dollars”), while existing UMG shareholders keep their shares in a reorganized Nevada corporation listed on the NYSE. The €40B implied valuation from the €22/share price (a 29% premium to the €17.105 close) balloons to €55B only by crediting Ackman’s activist plan — borrow €5.4B, sell UMG’s €1.5B Spotify stake, buy back ~17% of shares — as if those actions were already priced in.
The SPARC structure Ackman invented inverts the SPAC model: instead of raising money first and hunting for a target, a SPARC finds the target first and then asks rights-holders to fund it. This means the €1.1B from SPARC starts at zero and only materializes if investors opt in after seeing the UMG deal — making Ackman’s “we’re buying 100%” claim even more hypothetical, since a chunk of his equity check doesn’t exist yet.
Goldman Sachs Private Credit Corp. (GS Credit) narrowly stayed under the 5% quarterly redemption cap in Q1 2026, tendering for 17,285,147 shares (5.0%) and receiving requests for 17,281,858 (4.9990%) — clearing the cap by 3,289 shares. GS Credit touted this as proof of institutional-quality stability: peers Blue Owl (21.9% redemption requests), and Ares Strategic Income Fund (11.6%) all blew through the cap, with both attributing pressure to a small number of concentrated family offices and smaller institutions, not their broad retail bases.
Private non-traded BDCs redeem at 100% of NAV while publicly traded BDCs with similar underlying assets trade at 70–80 cents on the dollar — a pure arbitrage for anyone paying attention. The real question behind the redemption wave is whether it’s unsophisticated retail panicking (the bullish narrative for staying invested) or sophisticated money quietly exiting while retail holds the bag. Phil Blancato of Osaic, which manages $700B in client assets, captured the mood: a client called asking to liquidate “all my private credit investments” despite owning none.
JPMorgan, which spent much of last year desperately trying to buy private credit loans and failing to acquire a single one, is now trying to sell them. The bank is sounding out buyers for software loans from private credit funds seeking to reduce AI-disruption exposure — mostly at above 95 cents on the dollar. The reversal illustrates how fast sentiment shifted: the same illiquidity that made private credit attractive (no mark-to-market) is now its liability when sellers and buyers need to find each other.
A CEO spent his company’s Survivor-themed retreat bedridden with E. coli, listening through the wall as a hired ex-Navy SEAL ran drills on his team. His conclusion: “It sounds terrible out there, too.” The episode inadvertently validates the insight that shared misery builds teams better than manufactured fun — the chief product officer noted it generated “hundreds of inside jokes.”
Quotable:
“You get from €40 billion to €55 billion by counting the impact of your activist plans in the valuation of the company. You get from 6% to 100% by, essentially, syndicating most of the deal to the existing shareholders.” — Matt Levine on Ackman’s UMG deal math
The Core · email · 7 mins
On April 8, Trump announced a two-week US–Iran ceasefire on Truth Social, citing Iran’s “10-point proposal” as “a workable basis on which to negotiate.” Pakistan PM Shehbaz Sharif confirmed the truce covers Lebanon and all other theatres, effective immediately — a sudden pause to a conflict that began February 28 and has cascaded into a global energy crisis Fatih Birol (IEA Executive Director) called “more serious than the ones in 1973, 1979 and 2022 together.” The Strait of Hormuz blockade underpins every downstream disruption covered below.
India’s auto sector entered FY27 off a record high — 2.96 crore units in FY26, +13% YoY — but the West Asia conflict has already cut full-year FY27 wholesale growth forecasts from 8% to 3–6%. The conflict choked industrial gas supplies, forcing OEMs and component makers (foundries, casting, paint shops) back to captive coal and diesel; styrene and carbon black costs doubled, polyethylene prices surged 40–50% in Asia. FADA’s dealer survey found 53.2% reporting supply or dispatch disruptions, with 17.1% seeing delays of three or more weeks.
Price hikes are the auto sector’s immediate response: Maruti Suzuki has signalled a revision, and Tata Motors and Mahindra have already announced April hikes. Some Tier 2 component makers have suspended operations temporarily and are distributing induction cookers to workers to prevent labour flight — a sign of how deep the LPG shortage runs. The food economy is equally strained: commercial LPG shortages have cut menus at roadside eateries and sweet shops across India, while edible oil imports fell nearly 9% in March to 1.2 million tonnes.
Aviation is under compounding pressure. India cancelled 10,000 flights to West Asia since the conflict began, collapsing daily routes from 300–350 to just 80–90, while rerouted long-haul flights triggered DGCA to temporarily relax pilot flight duty limits by up to 1 hour 45 minutes — drawing safety objections from pilot groups. Air India simultaneously lost CEO Campbell Wilson after nearly four years, against a backdrop of persistent losses and regulatory scrutiny following a 2025 crash that killed 260 people; both Air India and IndiGo implemented new distance-based fuel surcharges effective April 8, ranging from Rs 299 domestically to $280 for North America/Australia routes.
Quotable:
“The oil and gas crisis triggered by the blockade of the Strait of Hormuz is more serious than the ones in 1973, 1979 and 2022 together.” — Fatih Birol, IEA Executive Director, to Le Figaro — invoking the Arab oil embargo, the Iranian revolution shock, and the post-Ukraine invasion crisis as the three benchmarks the current crisis exceeds
Ben Thompson · rss · 7 mins
Broadcom announced a deal to supply Anthropic with 3.5 gigawatts of TPU-based computing capacity starting in 2027, as part of a broader expansion of the Google-Broadcom collaboration that also includes a supply assurance agreement for Google’s own next-gen AI data center racks through 2031. Simultaneously, Anthropic disclosed that its annualized revenue run rate has surpassed $30 billion — up from ~$9 billion at end of 2025 — and that the number of business customers spending over $1M/year doubled from 500 to 1,000+ in under two months.
Dario Amodei has deliberately under-bought compute relative to the 10x annual revenue growth curve, because the math of over-committing is catastrophic: buying compute sized for $1T in revenue when actual revenue comes in at even $800B would be company-ending with no hedge available. His conservatism was explicitly about not “YOLOing” like competitors, favoring Anthropic’s enterprise revenue base (more stable, better margins) as a buffer — but with revenue now on track to 10x 2025’s run rate well before year-end, that conservatism is now becoming a binding growth constraint.
Back-of-envelope math on the 3.5 GW deal: at ~$50B/GW to build, total capital is ~$175B; using Google’s 6-year depreciation schedule gives ~$30B/year in depreciation costs; applying Google Cloud’s ~30% margin implies ~$43B in cloud revenue; applying Anthropic’s 40% gross margin implies ~$72B in Anthropic revenue. Rounding generously, the deal might cover ~$100B in incremental Anthropic revenue — likely not enough given the growth trajectory, suggesting more compute deals are coming.
Google is the natural and essentially only viable partner for this scale of compute: Epoch AI estimates Google controls the equivalent of ~5 million Nvidia H100 GPUs — more than any other hyperscaler — and unlike peers, Google’s fleet is dominated by custom TPUs rather than Nvidia chips. SemiAnalysis predicted in 2023 that Google’s TPUv5 buildout would give it more capacity than OpenAI, Meta, CoreWeave, Oracle, and Amazon combined, and the Epoch AI data confirms that prediction held.
Google’s strategic motivation is layered: it is spending $175–185B in capex in 2026 alone (the most of any tech company on AI), and cannot fill all that capacity with its own Gemini enterprise offerings. Routing Anthropic’s workloads through Google Cloud still captures AI compute revenue, just at lower margins than winning enterprise directly. More pointedly, every dollar Anthropic takes from OpenAI in enterprise weakens Google’s primary consumer-market rival, while Google’s massive capex advantage slowly spends OpenAI into the ground.
Quotable:
“Even though a part of my brain wonders if it’s going to keep growing 10x, I can’t buy $1 trillion a year of compute in 2027. If I’m just off by a year in that rate of growth, or if the growth rate is 5x a year instead of 10x a year, then you go bankrupt.” — Dario Amodei to Dwarkesh Patel, explaining Anthropic’s deliberate compute conservatism
The Information AM · email · 5 mins
Anthropic has topped $30 billion in annualized revenue as of April 7, 2026 — a 58% surge since just the end of February and more than a tripling from $9 billion at the end of 2025. The vast majority comes from API access to its models, with a smaller slice from Claude premium subscriptions.
Anthropic has now overtaken OpenAI on this metric: OpenAI was at $25 billion in annualized revenue in February. The gap has not only closed but reversed, with Anthropic running $5 billion ahead despite being the smaller, newer company.
Revenue growth is driving a compute arms race: Anthropic announced a new deal with Google and Broadcom for multiple gigawatts of Google TPU capacity starting 2027, on top of its existing infrastructure. CFO Krishna Rao framed it as “building the capacity necessary to serve the exponential growth we have seen in our customer base while also enabling Claude to define the frontier of AI development.”
Quotable:
“We are building the capacity necessary to serve the exponential growth we have seen in our customer base while also enabling Claude to define the frontier of AI development.” — Anthropic CFO Krishna Rao, April 7, 2026
CJ Gustafson from Mostly Metrics · email · 10 mins
Big tech is now acquiring distribution the way it used to acquire products. HubSpot bought Starter Story, Plaid bought This Week in Fintech, and OpenAI bought TBPN — the popular daily tech podcast — reportedly for “low hundreds of millions.” That last figure means OpenAI spent roughly 0.1% of its valuation to buy the trust and audience two podcast hosts built. Brands are treating newsletter and podcast audiences as balance sheet assets, not marketing line items.
Creator sponsorships work through a three-layer economic structure, not just ad buys. AG1 spends over $2M/month on podcast ads, pays Joe Rogan more than $10M/year across ads and affiliate fees, and layers on ~20% affiliate commissions plus whitelisting (running paid ads through the creator’s own account so they appear organic). The 20% affiliate cut on recurring monthly subscribers compounding over 3 years is a massive incentive — it turns casual mentions into active selling. Annual exclusivity deals also block competitors from the same channel entirely.
Free products and “perk passes” replace traditional paid acquisition. Kyle Poyar bundles year-long free product trials (from companies like Lovable) into his paid newsletter subscription. Companies get direct distribution to a curated, high-intent audience; their CAC is the product cost, not media spend. Crucially, Kyle takes no referral fee — the company wins on fit and conversion, not on bribed placement.
AI companies are booking customer acquisition directly to COGS by giving away inference credits. Cursor, Replit, and Lovable distribute free compute credits rather than running paid ads. Lovable takes it further — you can build a product without even creating an account, eliminating all friction before the “penny gap.” This shows up as heavy Cost of Goods Sold, not S&M, and only makes economic sense if the retained and expanded customer lifetime value dwarfs those initial free tokens.
Becoming the default inside an AI-native product is a new partnership moat. Lovable defaults new users to Supabase as their backend database. Vercel is pre-training Claude to recommend it as the best deployment tool for AI-generated code. The P&L implications are layered: ongoing rev share (negotiate one-time to protect unit economics), annual platform/exclusivity fees, well-paid partnership manager headcount, and a dedicated API integrations team — at PartsTech, where Gustafson was CFO, unannounced upstream API changes by suppliers like Napa Auto caused outages when the integrations team was caught off guard.
In-house media only works if the content is good enough that people would pay for it. Beehiiv’s own Creator Spotlight newsletter has 400K+ subscribers at a 40% open rate. Freightweaves built a trucking media brand and now sells SaaS into that audience. The failure mode is cheap, junior-written content nobody shares — you need a practitioner-level expert at ~$300K/year (or acquire a creator outright) to produce content that stands on its own merits and compounds into brand trust.
Community-as-distribution converts education directly into pipeline. GCAI, a legal AI company, teaches AI for Legal courses on Maven — a platform general counsel actually pay attention to. Nearly 25% of course takers converted to paying GCAI customers, an extraordinary ratio in a category (enterprise legal software) where cold outreach rarely lands.
Quotable:
“Their CAC was $80 to Dominoes. And their product was $10,000 a year.” — on a B2B SaaS company that cold-called auto garages after sending surprise pizzas, converting warm conversations into $10K ARR deals
Byrne @ The Diff · email · 7 mins
Two of the biggest planned IPOs of 2026 — SpaceX/xAI and OpenAI — both have serious management instability. All eleven co-founders of xAI who joined Musk three years ago have now left, an annualized churn rate of ~57%. OpenAI’s COO has been moved sideways to “special projects” and the head of product and business is on medical leave — the kind of turnover that would normally cause a company to delay going public.
High executive turnover at frontier AI companies is structurally inevitable, not a red flag. Senior roles at these companies have never existed before, org charts shift constantly as the same product pivots between revenue source and R&D cost center, and the incentives are inherently mismatched — outside executives import big-company pathologies, but neither side knows which pathology is harmful until it’s too late. The result: a ready-made excuse to cycle through senior leaders, with high opportunity cost making it worse to leave seats empty.
Right now is paradoxically the best moment to IPO with “hair on the deal.” A current cohort of retail investors actively prefers governance problems, conflicts of interest, and misaligned incentives — they treat drama as a feature. Companies contemplating IPOs should rationally factor this in and raise from that audience while the window is open, because cheap capital from entertainment-seeking investors is not a permanent feature of the market cycle.
Anthropic hit $30bn ARR, up from $9bn at year-end 2025 — faster to that milestone than Microsoft, Oracle, or Salesforce, all of which took at least 25 years. Meanwhile, labs collaborating on catching Chinese model distillation (OpenAI, Anthropic, Google sharing detection data) reveals the underlying motive: preserve premium pricing tiers against a price floor set by open-weight models.
Quotable:
“There’s a population of retail investors that likes governance problems, conflicts of interest, and misaligned incentives. It’s entirely rational for companies that are considering IPOs to incorporate this into their decision.” — on why messy AI companies should IPO now, not later
Big Think · email · 7 mins
The “grayening” is a documented, measurable trend: a 2020 Science Museum Group study fed ~7,000 photographs of everyday objects (kettles, lamps, cameras) spanning the late 1800s to 2020 into an algorithm and found a striking, long-term shift toward achromatic — neutral — colors in material culture, accelerating sharply in the 21st century.
Mass production industrialized blandness. Neutral tones are easier to standardize, less likely to clash across cultures, and more globally marketable than a bold hue like tangerine, which may sell brilliantly in Seville but offend buyers in Seoul. Gray became the muted lingua franca of global commerce — inoffensive precisely because it signals nothing.
Modernist ideology gave desaturation its philosophical backbone. Austrian architect Adolf Loos argued in his 1908 essay “Ornament and Crime” that color and ornamentation were signs of arrested moral development. His ideas shaped the Bauhaus school, which shaped the International Style of global architecture, and Le Corbusier doubled down by declaring color “suited to simple races, peasants and savages.”
Cars chart the collapse most precisely. The 1970s were the last bold decade — “Plum Crazy Purple,” “Hugger Orange,” “Lemon Twist.” By 2004, ~60% of new U.S. cars were achromatic (black, white, gray, silver); by 2023, that figure hit 80%. BASF’s 2025 Color Report shows white (38%), black (23%), gray (19%), and silver (8%) now account for 88% of all new cars globally. The most popular actual color — blue — is at just 6%.
Around 2010, minimalism became an aspirational identity — “Millennial gray” as conspicuous restraint. Apple’s iMac once came in Bondi Blue, Tangerine, Strawberry, and Grape; its current lineup is Space Gray, Silver, Gold, and Midnight Black. House paint names — Agreeable Gray, Mindful Gray, Accessible Beige — don’t describe a color so much as a social posture: Please, don’t mind me.
Gray correlates measurably with psychological distress. University of Manchester researchers asked healthy volunteers, anxious people, and depressed people to choose a color representing their mood: yellow dominated among healthy subjects; gray monopolized among both anxious and depressed participants, who described it as “a dark state of mind, a colorless and monotonous life.” Signs of reversal are now appearing — Pantone’s 2025 color of the year is Mocha Mousse (warm brown), Behr’s is Rumors (deep ruby red), green car sales doubled to 4% in the Americas in 2025, and Axalta named Evergreen Sprint its global car color of the year.
Quotable:
“Agreeable Gray doesn’t so much describe a color as a social posture.” — on how neutral paint names signal deference rather than identity
Amanda Natividad · email · 6 mins
Attribution and measurement are different jobs — and most teams only do one. Attribution assigns credit to the channel at the conversion finish line. Measurement asks whether a channel caused lift: more branded search, faster pipeline velocity, better conversion rates among people who already recognize you. Treating attribution as the whole story means you’re optimizing for demand capture while blind to demand creation.
76% of web attention lives where attribution can’t see. Rand Fishkin’s SparkToro research on the 5,000 most-visited websites found search accounts for only ~24% of visits — the other ~76% is social, newsletters, forums, Slack, and other destinations where opinions form and brand familiarity builds. The typical B2B buying journey: someone sees a LinkedIn post, reads a newsletter mention, gets a Slack link from a colleague, then Googles the brand name and converts. Attribution credits SEO (or SEM if there’s a paid result). The three moments that actually created demand don’t appear anywhere in the dashboard.
CMOs who shift budget from brand to search are defunding the thing that makes search work. When attribution data shows “search converts better,” the instinct is to pull spend from organic social and brand marketing toward Google Ads — but that eliminates the upstream influence that was warming buyers before they ever searched. The efficiency argument runs the other direction: when brand presence is strong and people arrive already familiar, CPAs drop, CTRs rise, and conversion rates improve because the ad is reminding someone of a brand they’ve already encountered, not introducing a stranger.
Quotable:
“Every time a CMO pulls budget from brand marketing or organic social to put more into Google Ads because ‘the numbers say search converts better,’ they’re defunding the activities that made search work in the first place.” — on the self-defeating loop of last-touch attribution
Lenny’s Newsletter · email · 7 mins
Embarrassment fear is the main reason creators quit, but it’s a false signal. Nobody remembers the projects that flopped or seemed cringe — audiences only retain the successes. Persistent creators who keep returning after pauses are the ones who end up with something genuinely good.
Optimizing for the algorithm is a trap: platforms change their ranking logic, and work made to please today’s feed looks baffling in hindsight. The durable filter is personal emotional resonance — make things that make you laugh, cry, or react strongly, because that specificity is what actually connects with other humans.
Wasting peak caffeine/focus on low-stakes email loops (“sounds good!”) depletes the mental alertness needed for real creative work, creating a defeat-and-refuel cycle that blocks progress. Guard the first hours of high alertness for the hard creative task; email can wait until the tank is lower.
Failed ideas aren’t dead ends — they’re a dormant pile to return to. Unworkable concepts sprout in new directions when revisited weeks or months later; the right move is to accumulate them rather than discard them. Rumination, by contrast, is pure waste: the only productive use of looking backward is extracting a single forward-facing improvement note.
Quotable:
“If people hate it, congrats, it’s popular enough to attract scorn. Maybe nobody sees it? Congrats, it’s a hidden gem.” — on releasing work and releasing control of its reception