New Ask Claude or ChatGPT about any stock, using Bullpen. Bullpen in Claude or ChatGPT. 7 days free
Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Friday, November 13, 2026

Back to the front page

From X

@StableBread

Fajasy on X

Rising memory chip prices are hitting Toast’s $TOST hardware business, so the restaurant software and payments company has been buying chips early. $TOST’s hardware and professional services gross loss widened to $85M in Q3 from $57M a year ago, and the $225M lost through September already tops the $220M lost in all of 2025. It doesn’t help that hardware revenue was flat at $45M while hardware costs rose 29% to $130M. Management has tied that pressure to rising memory chip prices, which are largely outside Toast’s control. What $TOST can control is when and how it buys, and that’s exactly what it’s been doing. It’s been buying memory chips early, which took inventory from $114M (as of December 31, 2025) to $244M (as of September 30), with another $191M of hardware purchases committed over the next 12 months. President and CFO Elena Gomez said on the Q3 call that “nothing has really changed” since the Q2 call, when she said Toast had already cut the expected memory cost for both 2026 and 2027 by using earlier hardware generations, cheaper memory, and spot-market buying. $TOST also started buying Toast Capital loans from its bank partner and holding them itself in 2026, buying $145M through September after none in the first nine months of 2025. Gomez said that lets Toast keep the full return instead of just a fee for arranging the loans. But Toast now also takes the loss when a restaurant doesn’t repay, just as the FTC challenges its practice of making the full loan due when a customer cancels. So far, Gomez says defaults are “within our expectations,” and most new loans will still go through Toast’s existing program rather than its own balance sheet.
@convequity

Convequity on X

Anthropic grew revenue 47% in the quarter OpenAI undercut it on price Anthropic's Q3 numbers in the IPO filing are best read against Q2. Revenue was $17.0bn, up 47% quarter on quarter, after 144% growth in Q2. The slowdown came as OpenAI cut prices against Anthropic with generous usage resets and extra subscriber perks. Holding 47% growth through that is a strong result. Q4 could re-accelerate. Opus 5.5 does much more engineering work per dollar than Opus 4, and it is better on other tasks too. Compute gross margin fell to 42% from 52%. Ten points is a smaller drop than many would expect, given what new capacity costs. My assumptions: capacity cost of about $35bn per GW in Q2 and about $50bn per GW for the latest AI data centers in Q3. Cost takes 48% of revenue at a 52% margin and 58% at a 42% margin. So revenue per GW rose from about $73bn ($35bn / 0.48) to about $86bn ($50bn / 0.58). Algorithmic gains offset a good part of the higher capacity cost. For next year I expect compute gross margin to keep drifting down slowly, perhaps a few more points. In my view, the higher cost of new capacity means more of the value goes to companies that can solve the energy bottleneck: SpaceX ($SPCX), and neo-clouds that have power, such as $IREN and Nebius. I will be tracking this as the roadshow starts this week.
@gnoble79

George Noble on X

From my friend Mr. X: As Anthropic apparently starts their road show, which I would call “We Work Part Deux.” Remember, as I said before, We Work cut their valuation 3x, still couldn’t get it done, and then eventually ended up bankrupt. And in my view, We Work and Adam Neumann had better governance and were a better deal, for a variety of reasons. And that was before everything started coming undone and prior to policy heading in the wrong direction and prior to the head of the central bank of NVDA, who is also financing them, was caught canoodling with a Chinese spy. This is why it does not work. This is not .COM or railroads. The $ are being spend on an incredibly dis-economic technology that will decay as opposed to sticking around. This is rapidly depreciating immature broken software and compute that is the equivalent of UNIX, SPARC, and VAX or a Mainframe. Then, the more they are spending, the less they are getting. It is sort of like the impact of every incremental dollar of debt on GDP, very diminishing marginal returns. There is no scaling. If the underlying technology does the opposite of scaling, the business cannot scale. Said simply, cost per unit is going way up (Anthropic’s S-1 tells you this too), revenue per unit is plummeting. See this tweet and subtweet: [] More on Jensen the China lobbyist: [] ICYMI re my last: $NVDA @JG_Nuke @DougKass @edzitron @MacrostrategyP @GaryMarcus
@StableBread

Fajasy on X

From $14.47/share, you’re risking a $4.66 loss (-32.2%) on Quantum $QMCO in my bear case for a $20.40 gain (+141.0%) in my bull case, before any fallout from the CFO firing, which none of the three cases counts. Here’s where $QMCO’s stock could be by mid-2027 in my bear, base, and bull cases. I value each case on FY2028 revenue (year ending March 2028), since the market will be pricing that year by then. Each case’s value is FY2028 revenue times the multiple, plus the $77.0M of cash, over 44.6M diluted shares (44.7M in the bull case). Bear case ($9.81, -32.2%) The bear case assumes $IBM’s drive supply stays flat, Toshiba’s added capacity ends the disk shortage that helped push customers toward tape, and the “temporary benefit from higher pricing” $QMCO got in the June quarter fades. That implies FY2028 revenue of $300M, 8.5% below the $328M annual run rate in September-quarter guidance, at 1.2x, the multiple at the $9.42 PIPE price, for a value of $9.81/share. Base case ($17.89, +23.6%) The base case assumes $IBM raises drive output gradually through 2027, letting $QMCO ship more of its backlog each quarter. That implies FY2028 revenue of $360M, 9.8% above the $328M run rate, at 2.0x, a discount to $HPE’s 2.3x TTM revenue (as of October 27) for $QMCO’s open material weaknesses, for a value of $17.89/share. I use $HPE rather than $IBM (4.2x), since most of $IBM’s revenue comes from software and consulting rather than hardware. Bull case ($34.87, +141.0%) The bull case assumes $IBM ramps drive output early in 2027, the backlog converts quickly, and more hyperscalers (the largest cloud and internet companies) buy $QMCO’s Scalar i7 tape library. For reference, CEO Hugues Meyrath said the i7 deal $QMCO won from a hyperscaler in Asia-Pacific in the June quarter was valued at “well over eight figures.” That implies FY2028 revenue of $400M, 22.0% above the $328M run rate, at 3.7x, the multiple at my August update, for a value of $34.87/share. Winning more hyperscalers would also make $QMCO even more dependent on a few customers. Its top five customers made up 27% of FY2026 revenue, up from 21% in FY2025, and one customer made up more than 10%, when none did in FY2025. But the takeaway is that even the bull case, with faster $IBM supply and August’s multiple, lands just 1.1% above the $34.50 peak.

Investor Letters

All letters

QXO Inc Earnings Update: Q3 2026

QXO's Q3 sales rose 79% to $4.9 billion after TopBuild and Kodiak acquisitions, while acquisition debt lifted interest expense to $163…

QXO $13B
SixSigmaCapital: Navigating the Financial Markets 2 clicks

Top Rebound Stocks for Nov’26

Texas Pacific Land's 50% selloff reflects faded data-center expectations, while its Permian royalty business generated $498 million of 2025…

TPL $19B
Rebound Capital 2 clicks

ASML ($ASML) Deep Dive

ASML monopolizes EUV lithography, a critical tool for leading-edge chips, with production capacity stretched through 2028 and customer…

ASML $591B
Jimmy's Journal 2 clicks

ON THE TAPE with Guy Adami

Guy Adami discusses Nvidia, Anthropic's IPO prospects, energy, gold and the AI trade.

NVDA $4T
WHAT ARE WE DOING? “Contrarians At The Gate” 5 clicks

DoorDash’s FBA Moment

DoorDash’s $17 billion Instacart deal targets DashMart warehouse scale, where 99%+ order accuracy and faster delivery address the limits of…

DASH $75BKR $41B
MBI Deep Dives 3 clicks

Publications we follow

Every publication on this site, hand-picked.

130 publications

Newsletters 110

YouTube 4

Blogs 15

Seeking Alpha 84 of 289

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

91 accounts
Nominate an account

Investor letters we follow

376 firms

Sectors

Ordered by how often our editor reaches for them.