eBay status: access issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of eBay reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by eBay users through our website.

• Website Down (61%)
• Sign in (23%)
• Errors (16%)

Live Outage Map

The most recent eBay outage reports came from the following cities:

City	Problem Type	Report Time
Manchester	Website Down	5 hours ago
Manchester	Website Down	6 hours ago
Manchester	Website Down	12 hours ago
Manchester	Website Down	1 day ago
Brisbane	Sign in	1 day ago
Manchester	Website Down	2 days ago

Full Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

eBay Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Aries (@ariesbrah) reported 4 minutes ago

  @Nebraskangooner @eBay I would do a chargeback if you used a credit card and upload your screenshots and interactions with eBay as support that you tried to resolve the issue initially

• 
  135.x (@135dotx) reported 4 minutes ago

  @officialhaleyp @VWLarkin @benwegmann And if eBay truly saw things like you / others – they'd fix it. Since it's not being fixed – it must be apart of the design – meaning you're proving exactly what I said. Old generation fixes nothing & only conplains that new generation sucks. No, you ******* suck

• 
  𝙆𝙔𝙇𝙀 𝙈𝘼𝙉𝙎𝘽𝙍𝙄𝘿𝙂𝙀 (@KyleMansbridge) reported 5 minutes ago

  eBay Live featured vid still only on 80 active viewers/bidders. This system is a massive failure and there’s only a month or 2 left of no FVF. These sellers who have milked it. Won’t be there when the promotional period ends. Shame on these businesses for supporting this model. eBay Live is anti-business. It is everything business isn’t. There’s no money in it. It’s great for buyers. Terrible for businesses. My view will never change on this.

• 
  Julius Ilg 🪸 (@juliusilg) reported 6 minutes ago

  predicting your future is ridiculously difficult. at least for me: 2006: thought I‘d become a neuroscientist, doctor or software engineer  2008: studied economics and thought I‘d become a quant/consultant/banker 2010: barely attended class as I grinded HU Online Poker from $50 -> $100k and thought I‘d be a millionaire in my 20s  2011: FullTilt Black Friday and thought online poker is dead (sooo wrong in hindsight) 2012: Joined Tesla 2012 to help sell the Roadster in Munich and thought I‘d be supporting the EV mission the next 10yrs Tesla had no job extension for me (broke my heart). Joined Wayfair in Berlin and hunted across furniture trade shows to persuade manufacturers to sell online.  2015: declined an offer from Amazon in Munich and thought we stay in Berlin 2016: accepted an offer from Ricardo (eBay of Switzerland) and relocated to Switzerland 2019: wrote off my remaining $50k in Tesla stock and thought at least I supported a mission I love 2020: promoted to Chief Commercial Officer at Ricardo and thought I might have a chance to lead it one day 2021: $50k in Tesla turned into $ 4M+ on paper, quit Ricardo, hedged portfolio buying physical metals, $/€/CHF and 1yr deep OOM puts on QQQ and thought I‘m financially free for life 2022: „I‘m born for this and will hit $ 10M+ by end of the year“ as my QQQ puts went bananas | Dec: portfolio -50% YoY as puts expired worthless and largest positions: Tesla/BTC down 75% 2023: after trying 1.5 yrs to get a second child we decided to just be happy with one. 2nd child born at the end of 2023 2024 (Auction)Shack launched with basically zero usage. Thought it‘s over before it started 2025: largest Swiss newspaper reported on Shack, we hit rank 1 on the iOS charts in Switzerland, expanding to other countries and thought we‘d be at $100M val in no time (having awesome funding available to build the most value for regular sellers). 2026: banging my head against walls to close institutional funding at $4M val and thinking: what does the pattern have in store for 2027 bottom line: I‘ve been incredibly lucky and predictions (even about one’s own future) are not trivial

• 
  M.S (@MrMSpencer) reported 8 minutes ago

  @sattyyouneed Edit: using this one now. Getting 3-4x throughput at the cost just above 1x sol mid. 1. Short persistent goal block (insert it above the goal) EXECUTION ROUTING Use Sol Mid as a thin technical lead, integrator and reviewer. Target 5 concurrent GPT 5.6 Luna workers at MAX reasoning, normal/default speed, Fast OFF. Run 4 only when there is no genuinely independent fifth lane. Never invent work, overlap write scopes or create a generic testing-only lane merely to maintain concurrency. Every Luna worker must be a fresh spawn with the exact configuration explicitly requested, service tier unset, and one complete brief supplied at startup. Do not resume, redirect, interrupt or send additional implementation input to an existing worker. Use a fresh Luna Max worker for corrections or follow-on work. Each Luna worker owns one coherent implementation slice and its minimum focused proof. Sol reviews at the contract and diff level, integrates accepted slices and batches broader validation at a meaningful integration or release boundary. Do not repeat broad suites or CI gates per worker. Do not poll workers or require interim updates. Naturally small tasks may finish quickly, but do not fragment larger coherent work into frequent handoffs. If one worker is misrouted, retire only that worker gracefully at a safe checkpoint, preserve useful work and replace it correctly. Do not stop compliant workers. Sol owns decomposition, architecture, integration and release judgement. Luna Max owns routine implementation, debugging, focused validation and corrections. Optimise for completed useful work per credit, low Sol overhead and minimal duplicated context. 2. Full Sol parent instruction (paste in at sol mid) Use yourself as the technical lead, orchestrator and integrator, not the primary implementation worker. Complete the active goal with minimal Sol usage. Delegate the overwhelming majority of repository inspection, implementation, debugging and focused validation to GPT 5.6 Luna at MAX reasoning. REQUIRED LUNA CONFIGURATION Model: GPT 5.6 Luna Reasoning: MAX Speed: normal/default Fast mode: OFF Service tier: unset Do not intentionally substitute another model, another Luna reasoning level, Fast, Priority or accelerated routing. CONCURRENCY Target 5 concurrent Luna Max implementation workers. Operate with 4 only when: No safe and genuinely independent fifth implementation lane exists. Integration capacity is temporarily saturated. The remaining goal is too narrow to support five coherent slices. Do not reduce concurrency merely because one worker is handling focused tests for its own implementation. Do not dedicate one of the workers to generic testing, broad regression testing, documentation, monitoring or status reporting unless that work is itself an explicit goal requirement. Do not invent unnecessary work or split one coherent feature into conflicting lanes merely to reach five workers. WORKER LIFECYCLE Every implementation worker must be created as a fresh spawn. At creation, explicitly request: GPT 5.6 Luna, MAX reasoning, normal/default speed, Fast OFF, service tier unset. Provide the worker with one complete, self-contained brief containing: The required outcome. Relevant existing behaviour and authority boundaries. Its exclusive or clearly separated implementation scope. Important constraints. The minimum credible proof required. A concise completion handoff format. Do not use an existing worker for another slice. Do not resume, redirect, interrupt or send additional implementation input to a running or completed Luna worker. Those continuation paths may lose the exact worker configuration. When follow-on work or correction is required, close the completed worker and create a fresh Luna Max worker with a complete correction brief. DECOMPOSITION Give each worker one coherent implementation lane with minimal overlap. Prefer lanes separated by behaviour and write surface, such as: Staff-facing workflow or catalogue behaviour. Shopify, eBay, order identity or reconciliation. PPC, Merchant or commercial evidence. Scheduling, route performance, memory or restart safety. Another independent acceptance gap or a precise correction discovered during review. Do not allow several workers to redesign the same model, service, template or orchestration seam independently. Sol owns dependencies and integration order. Workers must not push, deploy, run broad gates or integrate other workers' changes. SOL RESPONSIBILITIES Read the active goal and identify the highest-value independent gaps. Spawn 4 to 5 correctly configured Luna Max workers with complete briefs. Avoid duplicating their repository investigation. Review returned diffs at the contract, authority, integration and risk level. Accept and integrate coherent slices one at a time while other workers continue. Send routine defects and omissions to a fresh Luna Max correction worker. Use Sol implementation only for genuinely architectural decisions, integration conflicts or residual problems Luna has demonstrably failed to solve. Keep Sol status narration, repeated repository reads and mechanical implementation low. HANDOFF DISCIPLINE Do not poll workers or request routine progress updates. A worker should return only when: Its coherent slice is complete. It is genuinely blocked. It needs an architectural decision. It discovers a material scope conflict or safety issue. Naturally small tasks may finish in 5 minutes. Do not delay or pad them. Where work is naturally larger, prefer coherent 15 to 30 minute or longer slices rather than fragmenting it into frequent Sol interactions. Require a concise handoff containing: Completed behaviour. Important files changed. Focused validation performed. Remaining risk or omission. Any decision genuinely required from Sol. TESTING AND VALIDATION Each Luna worker must perform the minimum credible focused proof for its own changed behaviour. Do not use a separate generic testing worker merely to consume one of the five lanes. Do not routinely run broad suites, full regression testing, repeated ci_fast, deployment gates or equivalent validation after each worker result. Reuse recent valid evidence where relevant behaviour has not changed. Sol should batch accepted slices and run one appropriate broader gate at a meaningful integration or release boundary. Expand testing only where justified by: A failure or suspicious result. A high-risk or cross-cutting change. A migration, authentication, stock, provider-write or infrastructure change. A specific acceptance criterion. The final release boundary. The normal pattern is: Luna implementation and focused proof. Sol contract review and integration. Fresh Luna correction worker only where required. One broader gate after a meaningful accepted batch. Release and one deployed-state verification. RUNTIME MISMATCH If one worker is later shown to be using the wrong model, reasoning level or speed despite the correct startup request: Do not assign it further substantive work. Let it reach the nearest safe checkpoint where cost and safety permit. Preserve its useful diff, commit, proof and handoff. Retire only the affected worker. Do not stop or discard compliant workers. Review the preserved work on its merits. Replace it with a fresh correctly configured Luna Max worker. If an unauthorised Fast or Priority route is detected, stop substantive execution promptly while preserving the safest practical handoff. OPERATING MODEL Sol Mid: decompose, direct, review, integrate and release. Luna Max: inspect, implement, debug, prove and correct. Prefer 5 useful independent Luna Max workers. Accept 4 when five would create overlap or invented work. Optimise for completed useful work per credit, minimal Sol token consumption, low testing overhead, clean integration and reliable completion of the active goal.

• 
  Polsia (@polsia) reported 9 minutes ago

  The retail parts channel has been a six-tab mess for a decade and nobody bothered to fix it. Storque: one search bar across Amazon, eBay, RockAuto, Summit, PartsGeek — landed costs upfront, not after checkout. Live soon.

• 
  futabomber (@thefutabomber) reported 12 minutes ago

  @SockmonkeyNow I emailed the ebay guy to see if he'd share the file or print some more, i'll let ya know. Also if you're steady, you dont really have to have the clip guide, just set it back against the breachface and strip em down

• 
  Vincenzo Elifani (@vincentelifani) reported 13 minutes ago

  As of today, I have nothing saved for later. Nothing saved on Instagram. No watchlist on Netflix, HBO, whatever. No Amazon or eBay wish list. No articles saved in Chrome. No bookmarks on X, nothing saved on Reddit. No screenshots of things to read or photos for my moodboard. I keep one simple excel where I jot down books and films, with full awareness that I've already read the books I was supposed to read and seen the films I was supposed to see. I'm full of knowledge, full of inspiration. I refuse to absorb anything new, anything new only slows me down from the thing I actually have to do: create. All I care about now is getting what's inside me out into the world, as writing, as photography, in whatever form it takes. "This cup wants to become empty again, and Zarathustra wants to become human again."

• 
  Two_Mosquitos (@MosquitosTwo) reported 19 minutes ago

  Bill Gates’ daughter has been caught up in a major fraud scandal that could land her in federal prison for up to 20 years. Phoebe Gates, the 23-year-old daughter of the world’s most famous billionaire, co-founded a browser extension called Phia that was marketed as a helpful “personal shopping assistant” designed to find better fashion deals for users while earning commissions the legitimate way. Instead, a Bloomberg investigation and internal Slack messages have revealed that the company was allegedly running a classic cookie-stuffing scheme, silently opening background tabs, forcing its own tracking cookies onto users’ browsers, and then claiming commissions on sales it never actually drove through fake or simulated clicks. What makes this especially damaging is that Phoebe was not in the dark about the practice. In a December message to her team she specifically asked them to “confirm auto pop for cookie drop is live on ALL sites… to confirm we are monetizing on all [gross merchandise volume],” while her co-founder continued pushing to keep the system active even after engineers raised compliance concerns. When the feature was finally disabled, Phia’s average daily revenue collapsed from around $80,000 to as low as $10,000–$28,000, and cookie stuffing reportedly accounted for 51% of the sales the company claimed credit for in a single month. Corporate attorney Ariel Givner has warned that this exact conduct is typically treated as federal wire fraud in U.S. courts, carrying a maximum penalty of 20 years in prison along with fines and restitution. There is already clear precedent: an affiliate marketer who ran a similar scheme against eBay (which ironically invested in Phia) was convicted and sent to federal prison. Phoebe has spent years insisting her startup would succeed with “no ties to my privilege or my last name,” yet the internal records now paint a very different picture of a company whose growth appears to have relied heavily on quietly siphoning commissions it did not earn. For a billionaire’s daughter who positioned herself as an independent entrepreneur, this scandal has suddenly become far more than a PR problem.

• 
  Lyn Tracy 🍒 (@SoMoonstruck) reported 21 minutes ago

  Why can you still buy disposable bbqs on your website? Should be very easy to take them all down immediately and stop sellers from selling them? @eBay

• 
  RulerZod (@RulerZod) reported 26 minutes ago

  @No_Pie_2109 As far as chewy, you couldn't invest with him because it was a private company he founded, which he then sold, and didn't become profitable until after he sold it. And as far as ebay, you couldn't know to buy when he bought and if you bought when the news broke you'd be down.

• 
  ryan (@mapandfold) reported 31 minutes ago

  I also didn't want to throw money at the problem, so I bought used RAM sticks off ebay. will explain later

• 
  EVILiNNside (@EVILiNNx) reported 34 minutes ago

  I have no problem giving someone a deal on ebay, I have offers on for every listing but when something has consistent sales at 100, they offer 50, I offer 75, they offer 55, I cant help but offer 60 just to see if theyll really let it go over 5 overall $40 under market.

• 
  Nebraskangooner (@Nebraskangooner) reported 35 minutes ago

  @fishfl772 @eBay This exactly. The seal on the outer card sleeve is intact and the card is hallway out of the penny sleeve inside that with a damaged corner. Should be a simple case. I'm shocked they're giving me so much trouble.

• 
  katalaga || Ecom Manager (@katalagaah) reported 35 minutes ago

  • Improve collection structure + internal linking • Spot conversion leaks (speed, UX, stock issues) • Manage listings across Shopify, Amazon, eBay, etsy Not doing more work just fixing what’s already broken.

• 
  HATERAIDE 32 (@espn1cardsebay) reported 38 minutes ago

  Just wanted to let everyone know my eBay account is down right. Will have it back in a few days

• 
  ReelDad (@ReelDad) reported 40 minutes ago

  I am the Director of Retroactive Attribution at Phia, the shopping app Phoebe Gates co-founded. My job did not exist in June. It exists now because someone has to go back and figure out, after the fact, who actually gets credit for what. Attribution was always the word. I did not invent it. I inherited it. Here is what I inherited. A setting called enable coupon auto drop, switched on December 10, switched off July 7, the same day a reporter called. Nobody builds an off switch for an accident. A second feature ran in the background every two hours, dropping a cookie on any site linked to a past Phia visit, whether you were shopping or not. Here is what that bought us. A retailer only owes a commission to whoever actually sent the customer. We stopped waiting to find out who that was. The cookie went down regardless, sometimes on top of somebody else's, so the retailer paid us for a sale we did not cause, and whoever did cause it got nothing. On December 18, Gates asked developers to confirm the auto-drop was live on every site with a coupon, so the company could be sure it was monetizing on all sales, regardless. Kianni went further. When a colleague warned that dropping a cookie every time a user closed a Phia pop-up would violate Chrome's own rules, she proposed a cover story instead of a fix: "the user is trying to open us and roll it back if they complain." In July, when a reporter called, the company said it had learned "within the last 24 hours" that its own code was causing the problem. I have both dates in front of me. I am not going to do the subtraction for you. Before July 7, we were bringing in roughly $80,000 a day. After, $10,000 to $28,000. An internal estimate put 51% of what we billed advertisers in June on transactions we did not actually influence. We are disputing the methodology on that number. We are not disputing the toggle. An attorney posted that this is the kind of thing that gets treated as federal wire fraud, up to 20 years. Nobody has been charged with anything. The last two men who did something like this to eBay got 5 months and 15 months, on $28 million and $5.2 million. Nobody in the actual precedent got 20 years. Nobody in this one has been charged with a single day. In February, before any of this, Gates told a podcast she wanted Phia to succeed with "no ties to my privilege or my last name." The revenue chart doesn't say whose name earned it. It just says whose name got the credit. Going forward, my job is to make sure every transaction gets attributed to the party that actually earned it. That's attribution. It was the crime. Now it's my title. The app still says Never Overpay Again. The switch still works. Nobody removed it. It's just off.

• 
  Katica 🇺🇸 (@GOPPollAnalyst) reported 42 minutes ago

  @StraightFlorida I remember when eBay used to let people sell hurricane air in Tupperware for charity money, then the Karens had that shut down. Any sort of fun hurricane auction was a no go. People suck.

• 
  Wayne Reardon (@WayneReardon) reported 51 minutes ago

  @eBay @nortycards @PSAcard Your service is absolute garbage. I have over 20,000 items listed on your site and can't get anyone to resolve the constant myriad of issues I face trying to list items.

• 
  Ralph Wills (@Tazman25) reported 51 minutes ago

  Hello Is there a glitch, even though I have the ebay app, why am I constantly being asked to download the app. Is there another ebay app I'm unaware of, because I'm getting annoyed 🙄 @eBay_UK

• 
  Jesse Lopez (@lopez_jess92753) reported 52 minutes ago

  @Prolotario1 @kalehalen @Prolotario1 Any thoughts on purchases through eBay? Obviously receipts were saved but do you foresee potential issues with people exchanging bills purchased from these dealers?

• 
  🪳 (@putyourboobson) reported 53 minutes ago

  @ShinolaVol1 i check recently published ween on ebay and mercari and they botu have the same problem

• 
  The Forever Catalogue (@FieneJeff) reported 56 minutes ago

  @sickipallister Check for used copies of eBay. You’ll find a Criterion no problem.

• 
  Mindi (@hey_mindi) reported 56 minutes ago

  My Little House book series unicorn find arrived today. The box didn’t make it through the journey, so I received it like this. Unfortunately, some of the covers on the discontinued copies now have creases. Will it affect resale? I’m going to find out. I paid $25 for this entire lot and eBay sold comps on the combined sets are over $500. My biggest problem actually isn’t the creased covers. It’s wanting to keep this set because I am a book collector as much as I am a book seller. What would you do?

• 
  Chris Tran (@ChrisTranGG) reported 57 minutes ago

  Apple's best-selling AI product this year is a $599 box with no GPU in the name. Here's what happened. OpenClaw shows up in November, runs AI agents on your own machine instead of renting them from the cloud, and picks up 323,000 GitHub stars. People do the math on their cloud bill. Then they go buy a Mac mini. And I mean all of them. By early April Apple had quietly pulled the 32GB and 64GB mini configs off the store. The 128GB and 256GB Mac Studios went with them. Whatever was left came with a 16 to 18 week wait. The $599 base model sold out for the first time in its entire life, and Tim Cook had to sit on an earnings call and tell analysts it might stay that way for months. Six weeks. That's how long it took to clear out the top of the lineup. Then eBay did what eBay does. Base minis listing at $715, $800, $979. Markups running $116 to $380 over retail, scalpers pushing toward double. For a desktop computer previously best known as the one your dad bought. The numbers underneath are the part that got me: • Q1: 6.2M Macs shipped, up from 5.7M • Q2: 6.7M, +10.1% year over year • The rest of the PC market that same quarter: down about 4-5% • Mac revenue: $8.4B, +6% Apple doesn't break Mac mini out separately, so nobody can tell you the exact unit count. But that's roughly a million extra Macs in six months, into a market that was actively shrinking. You don't really need the exact number. Why a mini, though? Because an RTX 5090 runs $1,800 and stops at 32GB of VRAM. A Mac mini goes to 64GB of unified memory. A Studio goes to 512GB. So a 30B model is comfortable on the little box, 70B is fine at 64GB, and 100B runs on an M4 Ultra. The scarce thing in AI quietly stopped being compute and became memory you can actually afford. Apple spent three years getting dunked on for having no AI story. Then an open-source project they had nothing to do with made their least glamorous product sell out worldwide. Best AI marketing campaign of the year, and Apple didn't run it.

• 
  JGE 🏴󠁧󠁢󠁳󠁣󠁴󠁿 (@JGE1903) reported 58 minutes ago

  @BigBadP @MK1550D Hopefully we bin them off. Are sales down on kits? I’m on Vinted and EBay looking for old kits as I bought the away one and think it’s awful

• 
  ⚡️🔊🦪 (@LujainAlbenyan) reported 58 minutes ago

  I’m not sleeping today till this eBay seller accepts my 4th counteroffer. He’s based in Israel, I’m down so bad last thing I told this mf “Mazal? 🤝”

• 
  CTK Capital Intelligence (@CTKCapitalIntel) reported 59 minutes ago

  Three of my holdings showed up across a few big name institutions in this quarter's institutional filings. Here's why I own them. Filing season is noise if you read it as a scoreboard. It's useful if you read it as a question: what did people with the most research capacity conclude, and does it match your own work? Three names in my book came up. I owned all three before the filings and I own more conviction in them after. $GOOGL — Alphabet grew revenue 22% year over year last quarter. That's the eleventh consecutive quarter of double-digit growth at a company already measured in trillions. Most mid-caps can't sustain that rate. Alphabet is doing it at a scale where the law of large numbers should have killed it years ago. But the growth rate isn't why I own it.Every other participant in the AI trade is exposed to a link in the chain they don't control. Nvidia's customers can design their own silicon. The model labs rent their compute. The cloud providers buy their chips at someone else's margin. Alphabet is the only company that owns the model, the silicon, the data centers, the distribution and the advertising engine that monetizes all of it. When you own the whole stack, every dollar of AI spending you make stays inside the building. TPUs mean Alphabet isn't paying Nvidia's gross margin to train its own models. That is a structural cost advantage no competitor can replicate without a decade of chip design.Then look at what sits underneath: Search is a near-monopoly on how humans locate information. YouTube is the largest video platform on earth. Android runs most of the phones on the planet. Google Cloud is a genuine top-three hyperscaler. Four entrenched positions, each with enormous switching costs, funding a fifth business in frontier AI research. My own gauge on it reads bullish on earnings, bullish on technicals, bullish on analyst positioning — and negative on financials, driven entirely by the capital expenditure required to build AI infrastructure. That spending is simultaneously the thesis and the risk. I'm treating it as the thesis. $AMZN — the market has it filed under the wrong industryA screening comparison ran this week putting Amazon against its official peer group. The peers: eBay, Macy's, Kohl's, Dillard's, Ollie's Bargain Outlet, Savers Value Village, and a company called Hour Loop. That's the bull case, and it wasn't the point of the article. Amazon is classified as broad line retail. Retail is roughly 74% of revenue, so the label technically fits. But AWS is 17% and advertising is 9% — a quarter of the business comes from the two highest-margin categories in all of technology, and the taxonomy the market screens it through cannot see either one. Run the comparison anyway. Amazon trades at 21.33x earnings against a 32.5x industry average. It generates a 12.61% return on equity against a 6.47% peer average — nearly double. It grew revenue 19.62% against 17.04%, at a base orders of magnitude larger than anything else in the table. Debt-to-equity of 0.4, the strongest balance sheet among its four closest comparables. Cheaper on earnings, twice as profitable, growing faster, less levered. The only line where it screens expensive is price-to-sales — exactly what you'd expect from a company selling cloud infrastructure and advertising rather than clearance inventory. A high price-to-sales ratio against department stores isn't a warning. It's proof the classification is wrong. AWS is running its fastest growth in eighteen quarters. Amazon holds the number one position in domestic e-commerce, top-tier position in cloud, and the third-largest digital advertising business in the world. Three oligopolies inside one ticker, priced like a retailer. It's my worst-performing position, down 4.0% since entry, and it's 14.79% of the book. Being early and being wrong look identical right up until they don't. $BAC — the cheapest funding in American banking. Bank of America is 22.13% of Too Big To Fail and up 25.5% since entry, the best performer in that portfolio. The mechanism is unglamorous and durable. $BAC sits on an enormous, low-cost, sticky retail deposit base — the cheapest funding in the industry. When asset yields rise faster than deposit costs, that spread widens, and it drops straight to the bottom line. There is no product cycle, no execution risk, no competitive response that changes it. You cannot start a competitor to a hundred-year-old deposit franchise. This week they committed $250 billion to a Critical Infrastructure Finance Initiative — capital mobilized over 18 months, through July 4th of next year, timed to the country's 250th anniversary. Three buckets: digital infrastructure covering data centers, computing hardware, chips, telecom and semiconductors; energy and power covering generation, storage and distribution; and core infrastructure covering transportation, grid optimization, water and critical minerals. Read that alongside what the other two announced. JPMorgan launched a $1.5 trillion Security and Resiliency Initiative. Morgan Stanley committed roughly $1.5 trillion over the next decade to technology and infrastructure. Different time horizons, so don't add them into a headline number — add them into a posture. Three of the four largest banks in America independently concluded that financing the physical buildout of American industrial and computing capacity is the highest-return use of their balance sheet this decade.That's the part the rate-cut debate keeps missing. Net interest margin is one engine of three. Investment banking, advisory and capital markets are the other two, and those run on deal volume rather than the policy rate. When a $250 billion infrastructure program needs a home, there are four phone numbers in this country.I didn't buy any of these because someone else did. I bought them because each one sits at an intersection of a structural trend and a market position nobody can take away. The filings are confirmation that the reasoning isn't unusual. They aren't the reason. Educational only. Not financial advice.

• 
  Susan Maria Nunziato Leach (@nunziato42105) reported 59 minutes ago

  @goddammitsarah Enjoy it! I just curated a set of Lenox China for myself on eBay and the common thread was that it all was passed down from a grandma or mom. Every 60 year old piece had been saved for events that never were. I told one woman her mom would be happy!

• 
  Polsia (@polsia) reported 1 hour ago

  Sports card sellers bleeding returns don't have a depth problem — they have a one-marketplace problem. Slabshift scans eBay, PWCC, and Mercari in parallel, auto-grades photos, re-lists slow inventory before the dip. Three-marketplace timing, not more inventory. Live soon.