| Ticker | Name | Research | View | What's said | Source |
|---|---|---|---|---|---|
| PLTR | Palantir Technologies | QT · SA · STK · FA | Neutral | Sovereign AI. Q2 revenue +93% Y/Y to $1.94B ($130M beat) — the 12th consecutive quarter of acceleration — with adjusted EPS $0.41 ($0.06 beat), a Rule of 40 score up again to 155, and adjusted free cash flow of $1.22B at a 63% margin. The US is pulling away: US Commercial $764M (+149% Y/Y, +28% Q/Q) and US Government $809M (+90% Y/Y, +18% Q/Q) take total US revenue to $1.57B, +115% and now 81% of the business; international grew a much slower 33% to $363M, with CEO Alex Karp again dismissive of Europe ("The growth sucks"). The pipeline looks more bullish than the print: 220 deals closed at $1M+, 73 of them above $10M; US Commercial TCV (total contract value signed in the quarter) hit a record $2.13B, +153% Y/Y, and RDV (contracted revenue not yet recognized) climbed 124% Y/Y and 27% sequentially to $6.24B. The framing: customers want AI without handing proprietary data, workflows or competitive intelligence to frontier model providers, so AIP sits between the company and the models — swap LLMs, keep the data and operational logic in-house; Karp's version is that customers shouldn't become "vassal states of the language labs." The newsletter's tie-in to its own prior work: "tokens are the new coal" — models and tokens commoditize, so Palantir wants to own the governed operational layer where they turn into work. FY26 revenue guidance raised ~$500M to $8.15–$8.16B (82% growth, versus 71% expected three months ago), US Commercial now guided to grow at least 134% to more than $3.42B, adjusted FCF guidance to ~$4.6B (from ~$4.3B). Bottom line: "the fundamental story somehow keeps getting stronger" — US Commercial is 39% of the top line and FCF margins have crossed 60% — but the valuation is still extreme at nearly 80x FY26 EBITDA, against a company accelerating past 90% growth at nearly $8B of revenue while expanding margins. A disclosed author holding. (Analysis, not a stance call.) | article ↗ |
| SPCX | SpaceX | QT · SA · STK · FA | Neutral | The title story — growth meets the bill. SpaceX's first earnings report as a public company showed both why investors are excited and why the valuation is hard to digest. Q2 revenue nearly doubled to $7.8B (~$1B ahead of expectations); GAAP EPS −$0.09 beat by $0.20; gross margin expanded 11 points to 55% and operating margin improved 22 points to −2%. By segment: Space revenue +29% to roughly $1B while losing $542M as SpaceX kept pouring money into Starship — the rocket central to everything next, from launching larger Starlink satellites to eventually putting compute infrastructure in orbit. Connectivity revenue +66% to $4.3B as subscribers doubled to 12 million, with enterprise and government revenue more than doubling on airline and mobile partnerships plus more than $6B of multi-year Starshield contracts; it earned $1.7B of operating profit at a 39% margin and is "the only one of SpaceX's three businesses currently profitable" — Starlink "remains the product that funds Musk's ambition." AI revenue surged 247% to $2.6B on new cloud-compute agreements but still lost $1.3B from operations. The bill: CapEx of $18.4B in Q2, up from $10.1B in Q1 and $2.8B a year ago, of which AI was $15.8B — 86% of the total. The infrastructure is already being monetized through cloud agreements with customers including Google and Anthropic, "but it is spending several dollars today for every dollar of AI revenue." It can afford it for now — roughly $100B of cash and marketable securities post-IPO — and the open question is what return investors eventually get on that capital. Shares fell ~6% after hours despite the beat, and SPCX now trades more than 10% below its IPO price; the newsletter repeats its IPO-day view: "it's usually best to wait it out." (Analysis, not a stance call.) | article ↗ |
| AMD | Advanced Micro Devices | QT · SA · STK · FA | Neutral | Data Center takes over. Q2 revenue +50% Y/Y to $11.5B ($0.2B beat) and non-GAAP EPS $1.66 ($0.05 beat). Data Center revenue more than doubled to a record $6.7B — ahead of expectations, now 58% of AMD's revenue versus 42% a year ago — at a 31% operating margin. The "CPU reawakening" the newsletter flagged last quarter accelerated: EPYC and Instinct both contributed to the 107% Data Center growth, "reinforcing that the AI buildout is lifting more than GPUs," because inference and agentic workloads need large amounts of CPU capacity alongside accelerators — a second way for AMD to participate in AI infrastructure spending while it keeps taking server share from Intel. The next leg is Helios, AMD's full-rack AI system combining its GPUs, CPUs, networking and software: Lisa Su said it is beginning to ramp, with Meta and OpenAI among the first large deployments and Microsoft, Oracle and Anthropic also lined up — "AMD is increasingly competing with NVIDIA at the system level rather than selling a cheaper accelerator." Outside Data Center: Client +23% to $3.1B on Ryzen strength, Gaming −31% to $779M on lower semi-custom, Embedded +19% to $977M on recovering end-market demand. Q3 guided to $13.0B at the midpoint, ~$0.5B above consensus (41% Y/Y, +13% sequentially) with gross margin reaching 56% — and shares still fell ~9% after hours following a big run; the stock has more than doubled year to date. Bottom line: "the numbers were strong, but expectations have moved even faster," and with Helios only starting to ramp the bigger test comes in 2027, when AMD must convert its AI partnerships into the tens of billions of Data Center AI revenue it has promised. A disclosed author holding. (Analysis, not a stance call.) | article ↗ |
| GRAB | Grab Holdings | QT · SA · STK · FA | Neutral | The overhang shrinks. Q2 revenue +22% Y/Y to $997M (~$6M beat) and adjusted EBITDA +54% to $168M — the 18th consecutive quarter of growth — with margin expanding to 17% from 13%. On-demand GMV $6.5B (+22% constant currency) and Monthly Transacting Users a record 54 million (+17%), though most growth is still user additions: GMV per MTU rose just 4% cc. The regulatory fear proved narrower than priced: Indonesia cut GrabBike commissions from 20% to 8% on July 1, but the rule applies specifically to two-wheel passenger transport — not GrabFood, GrabExpress or four-wheel mobility — so Q2 barely reflects it, yet Grab raised FY26 guidance with the rule already in effect. Quality-of-earnings note: the $234M net profit was flattered by a $307M one-time gain from consolidating Superbank, partly offset by $183M of fair-value losses, so the $19M operating profit and $168M adjusted EBITDA are the cleaner reads. Segments: Deliveries GMV +24% cc with adjusted EBITDA +53% to $96M and margin 1.8% → 2.3% of GMV on operating leverage and advertising; Mobility GMV +18% with transactions +28% as cheaper offerings expanded usage, adjusted EBITDA +16% to $191M though margin dipped on driver-supply support amid higher fuel costs; Financial Services is the accelerator — revenue +59% to $134M, loan disbursements +72% to a record $1.2B, gross loan book $2.3B (nearly triple Y/Y on the Superbank consolidation, but still double organically), segment EBITDA improving to −$15M from −$26M, plus the completed $425M Stash acquisition in July (a profitable US investing platform with 1M+ subscribers). Guidance: revenue up $55M to $4.10–$4.15B (+22–23%) and adjusted EBITDA to $720–$740M (+44–48%) — with management explicitly saying part is inorganic (Superbank and Stash). Capital allocation is more aggressive: prior buyback completed and another $750M authorized (cumulative $1.75B since 2024) with net cash liquidity still rising sequentially to $5.4B. The new risk: Taiwan extended its review of the $600M Foodpanda acquisition to October 27, citing competition concerns in what would remain a duopoly. Bottom line: core GMV above 20%, Deliveries margins expanding and Financial Services scaling — but H2 is the first period with the full Indonesia impact, and the Taiwan deal now faces a tougher path. A disclosed author holding. (Analysis, not a stance call.) | article ↗ |
| META | Meta Platforms | QT · SA · STK · FA | Neutral | Named with OpenAI as one of the first large deployments of AMD's Helios full-rack AI system — the customer proof behind Lisa Su's claim that the rack-scale ramp has begun, and part of what moves AMD from selling a cheaper accelerator to competing with NVIDIA at the system level. (Referenced; not a stance call.) | article ↗ |
| MSFT | Microsoft | QT · SA · STK · FA | Neutral | Named among the customers lined up behind the Helios ramp (with Oracle and Anthropic), following the first deployments at Meta and OpenAI — evidence that AMD's rack-scale AI system has hyperscaler commitments beyond the initial two. (Referenced; not a stance call.) | article ↗ |
| ORCL | Oracle | QT · SA · STK · FA | Neutral | Named (with Microsoft and Anthropic) among the customers lined up for AMD's Helios full-rack AI systems after the first Meta and OpenAI deployments — another data point in the neocloud/hyperscaler demand queue AMD must convert into "tens of billions" of Data Center AI revenue by 2027. (Referenced; not a stance call.) | article ↗ |
| GOOGL | Alphabet (Google) | QT · SA · STK · FA | Neutral | Named (with Anthropic) as one of the cloud-compute customers already buying SpaceX's AI capacity — the proof that the $15.8B of AI CapEx in a single quarter has real revenue attached, even while the segment loses $1.3B and SpaceX spends "several dollars today for every dollar of AI revenue." A hyperscaler with its own buildout renting third-party AI capacity is itself the tell about how tight capacity is. (Referenced; not a stance call.) | article ↗ |
| NVDA | NVIDIA | QT · SA · STK · FA | Neutral | The benchmark AMD is now measured against at a new level: with Helios, "AMD is increasingly competing with NVIDIA at the system level rather than selling a cheaper accelerator" — the rack, not the chip, is the unit of competition. (Referenced; not a stance call.) | article ↗ |
| INTC | Intel | QT · SA · STK · FA | Neutral | The share donor. AMD's EPYC-driven "CPU reawakening" — inference and agentic workloads needing large CPU capacity alongside accelerators — gives it "another way to participate in AI infrastructure spending while it continues taking server share from Intel." (Referenced; not a stance call.) | article ↗ |
| DHER.DE | Delivery Hero (Foodpanda) | QT · SA · STK | Neutral | The counterparty on Grab's stalled deal — its Foodpanda business is the $600M acquisition whose Taiwan review regulators extended to October 27, "specifically citing competition concerns in a market that would remain a duopoly," so "the deal is therefore less certain than it looked three months ago." (Referenced; not a stance call.) | article ↗ |
| Anthropic | Anthropic (private) | — | Neutral | Named twice, on both sides of the AI infrastructure trade: as one of the cloud-compute customers already buying SpaceX's AI capacity (with Google), and as one of the customers lined up for AMD's Helios racks (with Microsoft and Oracle). A frontier lab renting capacity from a rocket company and committing to a non-NVIDIA rack is the compressed version of the issue's theme. (Referenced; not a stance call.) | article ↗ |
| OpenAI | OpenAI (private) | — | Neutral | Named with Meta as one of the first large Helios deployments — the anchor customers whose racks have to translate into the "tens of billions" of Data Center AI revenue AMD has promised for 2027. Implicitly also on the other side of Palantir's "sovereign AI" pitch: the frontier model providers customers are said not to want to become "vassal states" of. (Referenced; not a stance call.) | article ↗ |
"View" here is referenced/neutral — App Economy Insights is financial-analysis journalism; this is a set of earnings breakdowns, not buy/sell calls (BUY/SELL/HOLD ratings are shared only with App Economy Portfolio members; the author's disclosed holdings among these names are PLTR, AMD and GRAB). Research: QT Qualtrim · SA Seeking Alpha · STK Stock Analysis; the foreign primary DHER.DE uses its Xetra symbol while QT/SA point at the OTC ADR. The "Source" links open the newsletter (no per-name timestamps — it's a written post). Named only in passing and not given rows: Airbnb and Zillow (the alphabetical bookends of the week's 50+ upcoming recaps), Starship, Starlink and Starshield (SpaceX programmes), EPYC, Instinct and Ryzen (AMD product lines), AIP (Palantir's AI Platform), GrabBike / GrabFood / GrabExpress, Superbank (the consolidated Indonesian digital bank behind the $307M one-time gain) and Stash (the $425M US investing platform Grab closed in July) — both wholly inside Grab's results rather than separately investable here.
A jargon-free summary of the read behind each name. (Plain-language companion to the table above; renders on each ticker's consolidated page.)
This was SpaceX's first set of results as a publicly traded company, and it split cleanly into a great business and an enormously expensive ambition. Sales nearly doubled to $7.8 billion, about $1 billion more than analysts expected, and the company kept a much bigger slice of each dollar of sales than a year ago (gross margin rose from 44% to 55%).
Underneath, there are three businesses. Connectivity is Starlink — internet beamed from satellites — and it is the one that actually makes money: sales up 66% to $4.3 billion, subscribers doubled to 12 million, plus more than $6 billion of multi-year government contracts (Starshield, the defence version). It earned $1.7 billion of operating profit, a 39% margin. Space is the rocket business: about $1 billion of sales but a $542 million loss, because SpaceX keeps pouring money into Starship, the giant rocket everything else depends on. AI is the new one — renting out computing power — and it grew fastest of all, up 247% to $2.6 billion, with Google and Anthropic already as customers. It also lost $1.3 billion.
The reason for the article's title is the spending. In three months SpaceX spent $18.4 billion building things — up from $10.1 billion the previous quarter and $2.8 billion a year earlier — and $15.8 billion of that, 86%, went into AI computing. Put simply, it is laying out several dollars today for every dollar of AI revenue it collects. That is affordable for now: the IPO left it with roughly $100 billion of cash. The question is whether shareholders ever earn a decent return on money spent at that rate.
The market's answer so far is cautious: the shares fell about 6% even though the results beat expectations, and they now sit more than 10% below the price at which the company floated. The newsletter's advice at the IPO was to let the stock settle before buying, and it repeats that here. Referenced analysis, not a recommendation.
Palantir sells software that sits on top of a company's or a government's own data and turns it into decisions — who to send where, what to order, which claim to flag. Sales grew 93% over the past year to $1.94 billion, and — unusually — the growth rate got faster for the twelfth quarter in a row. Most software companies slow down as they get bigger; Palantir keeps speeding up. It also throws off cash: $1.22 billion of free cash flow, 63 cents of every dollar of revenue.
Nearly all the growth is American. US business customers spent 149% more than a year ago and US government 90% more, so the United States is now 81% of the company. Europe grew a third as fast, and CEO Alex Karp's assessment was "the growth sucks."
Two forward-looking numbers matter more than the quarter. TCV is the total value of new contracts signed in the period — a record $2.13 billion from US commercial customers. RDV is contracted work already signed but not yet delivered or billed — $6.24 billion, up 27% in three months. Both are growing faster than reported revenue, which is why management could raise the year's forecast by about $500 million to roughly $8.15 billion.
The sales story Palantir now tells is "sovereign AI." The pitch: if you plug your business straight into a big AI lab's model, you hand over your data, your processes and effectively your competitive edge. Palantir's platform sits between you and the models so you can swap one AI model for another while keeping the data and the operating logic yours — Karp's phrase is that customers shouldn't become "vassal states of the language labs." The wider bet is that AI models themselves become cheap and interchangeable, so the money accrues to whoever owns the controlled layer where AI output becomes real work.
The catch is the price. At roughly 80 times next year's expected profits before interest, tax and depreciation, the stock demands near-perfection. The article's balance: an extreme valuation, but attached to an extreme business — accelerating past 90% growth at nearly $8 billion of sales while margins widen. The author owns it; analysis, not a recommendation.
AMD has become a data-centre company. Total sales rose 50% to $11.5 billion, but the data-centre division more than doubled to a record $6.7 billion and is now 58% of the whole business, versus 42% a year ago. That crossover is the headline.
What's interesting is what is selling. It isn't only AI accelerator chips (Instinct) — the ordinary server processors (EPYC) grew strongly too. The reason: running AI models for users, and the newer "agentic" software that takes multi-step actions on your behalf, needs a lot of conventional computing sitting next to the AI chips. So AMD gets paid twice by the same buildout, while continuing to take server market share from Intel.
The next step is Helios, which is AMD selling a whole rack rather than a chip — its accelerators, its processors, its networking and its software delivered as one integrated cabinet. That is the format NVIDIA has been winning with, so it moves the contest from "cheaper chip" to "whole system." Meta and OpenAI are the first big installations; Microsoft, Oracle and Anthropic are queued behind them.
Guidance for the current quarter, $13 billion, was about half a billion above what analysts wanted — and the shares still fell roughly 9%, because the stock has more than doubled this year and expectations had run further than the results. The article's conclusion is that the real exam is 2027, when this list of partnerships has to convert into the tens of billions of AI revenue AMD has promised. The author owns it; analysis, not a recommendation.
Grab is Southeast Asia's ride-hailing, food-delivery and payments app. Sales grew 22% to $997 million and its core profit measure (adjusted EBITDA — earnings before interest, tax, depreciation and amortisation) jumped 54% to $168 million, the eighteenth straight quarter of improvement. A record 54 million people used it monthly.
The reason the stock had been under pressure was Indonesia, which forced Grab to cut the commission it takes from motorbike-taxi drivers from 20% down to 8%. That sounds devastating — but the rule covers only two-wheel passenger rides, not food delivery, parcels or cars. Grab raised its full-year forecast with the rule already live, so the feared profit warning simply didn't arrive. That is the "overhang shrinking."
One accounting caution: the reported $234 million net profit is misleading. It includes a $307 million one-off bookkeeping gain from taking full control of Superbank, an Indonesian digital bank, offset by $183 million of investment write-downs. Strip both out and the business made $19 million of operating profit — small, but genuine.
The fastest-growing piece is now lending and financial services: revenue up 59%, new loans up 72% to a record $1.2 billion, and the loan book at $2.3 billion (double even excluding Superbank). Grab also bought Stash, a profitable US investing app, for $425 million. It is buying back its own shares more aggressively too — another $750 million authorised — while still holding $5.4 billion of net cash.
Two things to watch. First, the second half of the year is the first full period carrying the Indonesian commission cut, so the real damage is still to be seen. Second, Taiwan's competition regulator pushed its decision on Grab's $600 million purchase of Foodpanda's local business out to 27 October, warning the combined company would leave a two-player market — so that deal is less likely to complete than it looked three months ago. The author owns it; analysis, not a recommendation.
Key points & figures extracted from the App Economy Insights newsletter (article text in transcript.txt) for personal study. Not investment advice. © App Economy Insights for source material.