I've published 500,000 words on how to analyze, value, and manage stocks. Here's the 50-question checklist I run through when evaluating a business (free Excel version included): ↓ 1️⃣ Business Model 1. Can I explain what the company does in one sentence? Is it within my circle of competence? 2. Do I understand the unit economics and what drives profitability? 3. Will the business be around and stronger in 10 years? 4. Which parts of the business have operating leverage? 2️⃣ Financial Health 5. Can the company cover interest payments comfortably? 6. Is debt manageable relative to cash generation? 7. When does debt mature and are there refinancing risks? 8. Any accounting red flags, including off-balance sheet items or recurring "one-time" charges? 9. Is working capital a source or use of cash as the company grows? 3️⃣ Cash Flow Quality 10. Does operating cash flow consistently exceed reported earnings? 11. Is FCF positive and does it fully cover dividends, buybacks, and acquisitions? 12. What percentage of capex is maintenance vs. growth? 13. Are gross, operating, and FCF margins stable or improving? 14. Does the company earn high returns on invested capital? 4️⃣ Economic Moat 15. What specific moats protect this business? 16. Is the moat widening or narrowing over time? 17. Can they raise prices without losing customers? 18. Would it take competitors years and significant capital to replicate this position? 19. Do network effects or feedback loops strengthen the business over time? 20. Are substitute products becoming more or less competitive? 5️⃣ Management Quality 21. Do insiders/founders own meaningful equity? 22. Are insiders buying or selling significant shares recently? 23. Has management met past guidance and acquisition promises? 24. Are executives paid for per-share value or absolute size metrics? 25. Do share buybacks occur at attractive prices and is the share count stable or decreasing? 26. Does management communicate transparently in tough times? 6️⃣ Revenue Quality 27. Is revenue diversified with no customer or supplier concentration risks? 28. Do customers have high retention rates and switching costs? 29. Is revenue recurring or does it need constant reselling? 30. Is the cash conversion cycle improving? 31. Can the business grow without heavy marketing spend? 7️⃣ Growth Potential 32. Is there runway for years of growth in current markets? 33. Can the company reinvest at high incremental returns? 34. Will growth require dilutive external financing or increased leverage? 35. Has revenue and book value per share grown consistently? 36. Is growth primarily organic rather than through acquisitions? 8️⃣ Market Position 37. Is market share stable or growing vs. competitors? 38. Is the industry growing and can the company grow faster? 39. Could technology or new entrants disrupt this model? 40. Is the TAM realistic given all competitors' targets? 41. What external shocks is the business unavoidably tied to? 9️⃣ Valuation 42. What growth rate does current valuation require? 43. Is valuation low vs. history and comparable companies? 44. How many years of current FCF would it take to pay off the market cap? 45. Is the stock cheap due to temporary issues or permanent problems? 46. Does earnings yield exceed bond yields meaningfully? 47. Is there a specific catalyst for revaluation? 🔟 Risk Assessment 48. How would this company perform in a recession? 49. Have I thoroughly studied the bear case? What could permanently impair this business? 50. Do I have sufficient margin of safety for the risks identified?

$QXO lost $227M last quarter. Strip out the deal accounting and the real business is near break even. Q1 2026: → Net sales: $1,730.2M → Gross profit: $409.3M (23.7% margin) → Net loss: $(227.1)M → Adjusted EBITDA: $1.2M → Loss per share: $(0.35) The gap between the $227M GAAP loss and near-zero EBITDA is the whole story of an early-stage roll-up. Most of it's amortization ($116.9M, mostly acquired intangibles), plus depreciation, stock comp, interest, and one-time deal costs. Non-cash or deal-related. What's left is a core roofing business that's barely EBITDA-positive right now, because integration spending and a soft housing market are landing at the same time. And Kodiak and TopBuild aren't in the numbers yet. One real signal underneath: gross margin has climbed every quarter since the Beacon close, 21.1% --> 23.3% --> 24.2% --> 23.7%. Some is accounting noise, but on $8.6B of revenue each point is ~$86M. Don't grade QXO on this income statement. Grade it once Kodiak and TopBuild hit the P&L.

$BCAR is trading even closer to its trust value floor since my writeup on May 8th. Next best opp to $BRUN (prev $WLAC) pre-merger if you understand the mechanics and have the risk appetite: newsletter.stablebread.com/p…

In under two years, QXO $QXO went from a tiny software shell to $8.6B in revenue. It didn't grow there. It bought its way there. It raised $5B at launch, which Bloomberg called the largest equity offering the building products sector had ever seen, then went shopping: → Beacon Roofing: $11B (April 2025), the platform that made $QXO the largest publicly traded roofing and waterproofing distributor in North America. → Kodiak: $2.25B (April 2026), a $2.4B-revenue distributor of lumber, trusses, windows, and doors. → TopBuild $BLD: $17B (announced April 2026), set to roughly double EBITDA. TopBuild is the largest insulation distributor and installer in North America, an 18% EBITDA margin business versus the 8% QXO's roofing runs. Combined, the company is $18.1B of revenue, $2B of EBITDA, ~1,150 branches, and 28,000 employees, with #1 or #2 positions across insulation, roofing, waterproofing, and lumber. And the two he walked away from say as much as the wins: → Rexel: $9.4B hostile bid the board rejected as too low. He walked. → GMS: $5B bid, then he stepped aside when Home Depot outbid him at $110/share. Fast and aggressive, but he won't chase a deal past his number. The TopBuild deal still has to close in Q3, and it's already drawn the usual merger-objection lawsuits, including a Delaware suit seeking to delay the QXO shareholder vote. These almost always settle for added disclosure, but the close is the catalyst to watch. P.S. I wrote a 5k word deep dive on $QXO last week. Link in bio!

Once data needs to travel between server racks (which hold multiple servers), copper wires can't keep up. Copper signals only carry about a meter at AI speeds before they degrade. Past that, you need fiber optic cables. Optical transceivers convert the electrical signal into light, send it through the fiber, then convert it back at the far end. Every ~18 months, transceiver speeds double. 400G → 800G → 1.6T → 3.2T. Each new generation grows the total market because data volumes grow faster than prices fall. Coherent $COHR and Lumentum $LITE supply the photonic components. Fabrinet $FN does the high-precision assembly. As of 2025, Fabrinet held 100% share of 1.6T transceivers for NVIDIA's Blackwell platform. Innolight $300308.SZ is the Chinese leader.

Once you have the chip, you need to make 8 GPUs act like 1. Each chip in a server only works on part of the AI training problem. They share data constantly at terabit speeds. $NVDA's solution is called NVLink NVSwitch. It's a high-speed communication fabric (essentially the highway between chips) that lets 8 GPUs work as a unified system. This is one of NVIDIA's quieter moats. Whoever owns the communication protocol gets to price the silicon that plugs into it. Broadcom $AVGO and Marvell $MRVL sell competing networking chips that hyperscalers can use to build their own fabric. Astera Labs $ALAB makes connectivity chips that keep signals clean at the speeds AI requires. If these alternatives gain ground, NVIDIA's networking advantage erodes even if it keeps GPU share.

You've got the chips. You've got the memory. But connecting them at the speeds required is its own manufacturing challenge. The dominant technology is $TSM's CoWoS (Chip-on-Wafer-on-Substrate). Think of it as the manufacturing process that physically wires the chip and memory together into one finished product. Through most of 2024 and 2025 this was the actual bottleneck on AI chip output. Wafer supply existed, but CoWoS capacity didn't. $TSM is scaling CoWoS from ~35K wafers/month in late 2024 to ~130K/month by end of 2026. Still sold out through 2026, with $NVDA reportedly reserving most of the available capacity. Samsung Foundry and Intel Foundry are coming up as second sources. ASE $ASX and Amkor $AMKR handle older-generation packaging.

Next, the memory wall. Compute is useless without fast memory right next to it. Modern AI chips are bonded to 4-8 stacks of high-speed memory called HBM (High Bandwidth Memory). Without enough memory bandwidth, the chip sits idle waiting for data. This became the binding constraint in late 2023. SK Hynix $000660.KS qualified HBM3 with $NVDA first and ran with it. Q1 2026: 72% operating margin (all-time high) on ~$38B in revenue. Management said cumulative HBM commitments now cover more than three years of supply. Samsung $005930.KS spent two years improving manufacturing quality after falling behind on HBM3. Their next-generation memory (HBM4) is ramping for NVIDIA's Vera Rubin platform in H2 2026. Micron $MU is the third entrant and the only US-based supplier. Memory has always been a cyclical business in semis. Margins this high will eventually attract enough new capacity to bring prices back down.

Start with the chips themselves. An AI training server is built around 8 accelerator chips (basically GPUs designed for AI math) wired together on one board. $NVDA owns this layer. They've dominated for two years, but that's starting to change. Hyperscalers (AWS, Google, Meta, Microsoft) got tired of paying NVIDIA's premium and started designing their own chips. They partner with Broadcom $AVGO and Marvell $MRVL to manufacture them. $GOOGL's Ironwood TPU, $META's MTIA, Anthropic's training chips, and OpenAI's chip are all on Broadcom's roster. $AMD ships MI300/MI400 and has emerged as the credible #2 to NVIDIA. $MSFT, $META, and $ORCL have deployed it for inference workloads and some training. $INTC ships Gaudi but has struggled to gain hyperscaler adoption.

AI data centers, CPOs, memory, and chips are cool and all… but what about the eggs you had for breakfast? 🥚 Cal-Maine Foods $CALM is the largest egg producer in the U.S., selling about one in four eggs Americans eat. At ~$77/share, it trades at 5x trailing earnings, with a ~14% distribution yield (divs buybacks), $1.15B net cash, and no debt. That's a net cash floor of $24/share, ~31% of the market cap. Plus the company keeps on acquiring smaller producers, and is now pushing into prepared foods, egg patties, omelets, waffles, and pancakes, to become more than just an egg company. And with the company down 34% from recent highs, is the market punishing it too harshly just because egg prices came back down? Publishing a full writeup on the topic in 2 hours. Link in bio!

Clover, not Toast $TOST, is the share leader in small-restaurant POS. Toast's "roughly 1 in 5 U.S. restaurants" stat is real. But it measures Toast against every restaurant in the country, not the smaller-restaurant pool where Toast, Clover, and Square actually compete. A January 2026 Baird report estimated card-processing share for U.S. restaurants below the top 250 chains, ~75% of a ~$1.1T market: → Clover (Fiserv): 20%, 175,000 locations. → $TOST: 17%, 145,000. → Square (Block): 13%. → Global Payments: 11%. The ~171K locations $TOST reports is its total across all segments, enterprise included. In the below-top-250 slice, Baird credits it with ~145K. Clover is ahead by ~30K. Square is the one moving. Its food-and-beverage GPV grew 16% y/y in Q4 2025, slower than $TOST's 22%, but its parent Block said fine-dining and other full-service operators (sit-down restaurants) are increasingly switching to Square. Full-service is the segment $TOST leads. Then there's DoorDash. It hasn't launched a POS, but analysts report it's testing one in San Francisco, Phoenix, and New York. A projection reported by PYMNTS has DoorDash reaching 20% of U.S. restaurants by 2035 if it launches. Small restaurants already lean on DoorDash for delivery, so a combined POS-and-delivery offer would be hard to ignore. However, these threats are unlikely to impact $TOST's numbers in the near term. $TOST's win rates rose y/y in FY2025, and it's adding 30,000 net locations a year. Baird also surveyed 37 restaurateurs, and most said they liked their vendor and planned to expand with it, with no talk of leaving. So bulls aren't wrong that $TOST scaled. They're wrong if they think SMB leadership is settled. Clover is ahead in the SMB slice by Baird's count, and Square is pushing into full-service, where $TOST leads.