Dolphin Research

SpaceX 2Q26 First Take: This was SpaceX's first earnings release, and results were solid. Revenue continued to beat expectations, while losses narrowed faster, putting breakeven within sight. By segment:

1) Launch services: third-party launch volumes increased, with ASP and margins rising against the tide. Starship IFT-13 performed well overall.

Launch revenue was $960 mn, above the $870 mn consensus and up 29% YoY, driven by higher launch services revenue. External Falcon 9 commercial launches reached 10 in the quarter, vs. 7 in Q1 and 9 a year ago, as large customers added missions.

Per-launch pricing also rose, with implied ASP up from ~$54 mn last year to ~$65 mn. The increase primarily reflects a more favorable customer mix, with a higher share from large accounts.

On profitability, GPM improved by ~1,100 bps QoQ to ~66%. Despite expensing Starship test-flight costs (incl. hardware loss) through the launch P&L, margins still rose, supported by higher ASPs. This underscores SpaceX's ability to extract premium pricing given its dominant position and the lack of mature, low-cost alternatives.

On Starship progress, IFT-13 in Jul deployed the first batch of V3 satellites, validated Raptor in-space relight, and captured critical thermal protection system data. The booster was lost in a hard landing after partial engine relight failure, but as the second landing attempt for a V3 booster, the outcome was within a controllable iteration range.

2) Starlink: revenue topped expectations on stronger B2B/Gov., while user growth was broadly in line.

Starlink revenue was $4.29 bn, slightly above the $3.88 bn consensus, on B2B/Gov. strength. SpaceX has signed and activated in-flight Wi-Fi agreements with multiple airlines, and its Starshield unit secured a multi-year U.S. Gov. contract worth over $6 bn.

On B2C, Starlink users reached 12 mn, roughly doubling YoY. Net adds were ~1.7 mn QoQ, broadly in line with expectations.

Pricing remained stable (ARPU $66/month). In mobile, SpaceX secured rights to 65 MHz of spectrum in the U.S. (the former EchoStar 2 GHz band) and several global MSS licenses. While early-stage, SpaceX has effectively declared war on incumbent telcos, planning to launch V2 mobile satellites next year on the EchoStar 65 MHz band and build a terrestrial network to take share from AT&T, Verizon, and T-Mobile.

3) AI: revenue beat, with a sharp margin improvement.

AI revenue was $2.56 bn, above the $2.08 bn consensus, mainly as the compute lease with Anthropic (monthly fee $1.25 bn) began contributing. Margins improved meaningfully (GPM ~57%; OPM ~-49% vs. -115% expected), driven by premium pricing amid severe compute shortages. We estimate lease rates at ~$30–50 bn per GW, far above peers at ~$10–15 bn per GW.

Note that AI leasing contracts include a 90-day termination clause exercisable by either party. This creates inherent uncertainty in revenue visibility despite current strength.

Big picture, SpaceX's equity story resembles Tesla's: near-term performance is highly event-driven rather than earnings-driven, while a major lockup expiry in Aug 2026 could pressure the stock. The path forward hinges on key milestones: full Starship reusability, first mass launch of V3 satellites, whether Grok 5 exceeds expectations, and sustained signing of AI compute lease orders. $SpaceX(SPCX.US) $GS 2x Long SpaceX Dl ETF(SPAL.US) $SPCX 2X Short ETF(SNK.US)

AMD 2Q26 First Take: results broadly in line with the Street. Revenue growth re-accelerated, led by a doubling in the Data Center biz.

Data Center remains the key focus. Dolphin Research estimates AI GPUs (Instinct) delivered approx. $2.7bn this quarter vs. $1.1bn a year ago, with China export controls impacting roughly $0.7–0.8bn in the prior-year base.

For next quarter, the company guides revenue to $12.7–13.3bn, +10–15% QoQ (vs. the Street at $12.6bn). It expects non-GAAP GPM around 56% (in line with consensus).

With other segments relatively stable, we estimate next quarter’s incremental growth will mainly come from Data Center at roughly $1.2–1.8bn (vs. about $1.0bn QoQ this quarter). With MI455X starting to ship, that step-up is not particularly strong.

Investors care more about MI455X progress and the runway from here. Starting with MI455X, the company is shifting from selling chips to delivering rack-scale clusters, which lifts expectations for the AI GPU biz, yet the sequential guide still looks relatively soft.

AMD has announced collaborations with OpenAI, Meta, Anthropic, and Microsoft. The market reads this as growing customer acceptance of MI455X and is willing to ascribe a higher multiple on the prospect of faster growth.

Beyond signing wins, investors also want tangible execution, such as a more upbeat outlook, to meet elevated expectations. For more, follow Dolphin Research’s follow-up takes and Trans. $AMD(AMD.US)

SPOT 2Q26 First Take: Q2 was mixed, with headline results broadly in line.

(1) Price hikes are kicking in: Early-year increases in core US/EU markets started to flow through, with ARPPU up 6% YoY.

From Q3 guidance, the benign pricing lift should continue to support revenue.

(2) Ads are stabilizing: Q2 ad revenue was 450 mn, up 1% YoY.

On one hand, the FX headwind has rolled off, removing a passive drag.

In addition, starting this year, ad revenue excludes portions not directly tied to advertising, such as podcast subscription bundles, which have been reclassified into subscriptions.

(3) GPM is not keeping pace with pricing: Q2 GPM ticked up QoQ, with margin improvement in both subscriptions and ads.

However, management guides Q3 GPM down 30 bps QoQ.

(4) Operating investment is steady: Q2 cost control remained focused on R&D, likely via headcount efficiency.

Overall, near-term catalysts are limited.

Palantir 2Q26 First Take: Q2 results offset the Q1 slowdown in US commercial growth and kept the acceleration intact.

1) Guidance raised again: Q2 revenue rose 93% with further acceleration.

While the full-year guide implies a slower 2H growth pace, this is not the first time.

2) US commercial flexes again: Dependence on the US market climbed to 81% in Q2, driven by both expanding Gov. demand (reflecting broad Maven penetration across US agencies) and still-robust enterprise demand.

Management did not raise the 120%+ full-year guide for US commercial, implying 2H growth would slow to 68%.

3) Leading indicators remain healthy: Across TCV, RPO, RDV and customer count, momentum held up and both adoption breadth and depth improved.

(1) TCV represents total contract value of new bookings in the period.

(2) The key difference between RPO and RDV is cancelability, with RDV being broader and including cancellable orders.

(3) Customer adds were also net positive in Q2, led by enterprise clients.

4) GPM improved steadily with tighter cost control: Q2 GPM expanded 400bps YoY and was stable QoQ.

Beyond that, growth in the three opex lines was notably below revenue expansion, while SBC also contracted.

Luckin Coffee 2Q26 First Take: Overall, Q2 was solid despite last year's delivery-subsidy high base. SSSG turned negative to -5.3%, but store openings and user growth drove scale, with revenue up 28.5% YoY and Non-GAAP OP up 26.5%. Notably, delivery expense declined YoY for the first time since the delivery price war, beating expectations.

On a high base, revenue growth slowed QoQ. Q2 revenue was RMB 15.89bn (+28.5% YoY), beating the RMB 15.43bn consensus.

SSSG turned negative to -5.3% as expected. Monthly transacting customers reached 113mn (+23% YoY), a record above the Q3 2025 subsidy peak, suggesting users lost as subsidies faded were replaced by new-store additions and organic growth.

Store expansion accelerated rather than slowed. Luckin added 2,714 net stores in Q2, ending with 36,310 locations (+8.1% QoQ).

Self-operated units continued to outpace franchised stores, indicating ongoing densification in higher-tier cities. Overseas stores reached 223 with 46 net adds, with Malaysia the main contributor.

The marketing ratio edged up. Q2 GPM was 61.5%, down 130bps YoY, likely due to a higher mix of lower-margin raw-material wholesale.

Notably, delivery expense was RMB 1.62bn, down 3.1% YoY—the first YoY decline in absolute terms since the delivery price war began. As a % of revenue, it fell from 13.5% to 10.2%.

Store-level margin was 21.3%, just 20bps below last year. Despite SSSG at -5.3% and store count up 38.6% YoY, unit economics were largely intact, indicating continued efficiency gains at the store level.

In addition, the bottled RTD business formally launched in Q2. The company stepped up related spending, lifting the sales expense ratio by 100bps YoY to 5.8%, while G&A stayed flat.

Non-GAAP OP was RMB 2.396bn (+26.5% YoY). For more details, please follow Dolphin Research's detailed take and call transcript.$Luckin Coffee(LKNCY.US)

Amazon 2Q26 First Take: Heading into the print, Dolphin Research expected one thing: with Anthropic's ARR, born on AWS, rocketing, AWS should post a strong acceleration while retail just needs not to drag.

Key metrics:

a. AWS hit the fast lane: revenue of $42.2bn, growth accelerated 9ppt to 37%, an impressive pickup at scale; backlog reached ~$469bn, up 140% YoY.

More importantly, unlike many AI-driven cloud ramps that dilute margins, AWS expanded OPM by nearly 200bps QoQ to 39.4% this quarter.

b. Capex: $53bn in the quarter, vs. operating cash inflow of $45.4bn, driving FCF to -$7.7bn as operating cash can no longer fund the spend.

d. Retail: expectations were for 'no drag,' and results beat that bar by a good margin; Intl retail was softer, but its small base limits impact.

Overall, vs. Microsoft, Amazon's main blemish is that capex has outstripped operating cash flow.

a. Facilities: buildings have ~30-year lives, covering 5–6 server generations, so the annualized burden is modest.

b. Servers and networking: heavy ticket items, but procurement is timed close to go-live—only months in advance—so orders can be throttled if demand softens.

c. The company is willing to spend because it holds firm new orders in hand, allowing rapid revenue conversion as capacity comes online.

Payback on these investments is roughly 2–3 years, while servers and related gear have 5–6 years of useful life.

This mirrors our prior take on Google: invest against contracted demand to de-risk utilization, and the logic is sound.

Big picture, cloud is now accelerating sharply.

After reviewing several mega-cap prints, our focus shifts: as capex enters the heavy server and networking phase, cloud capacity will unlock quickly.

RBLX 2Q26 First Take: Under tighter kids' safeguards, Roblox stumbled again in Q2. It also took a self-inflicted hit in hopes of a future reset.

1) Self-imposed fixes weighed on engagement: DAU and quarterly hours came in below expectations due to age-rating compliance work in Q1–Q2. Russia and Turkey relaunched post age-gating, adding several million users. But Q2 metrics did not fully capture that recovery.

Beyond age gating, the platform also tweaked its algorithms. That sharply reduced short-term boosts and further pressured near-term results:

(1) Chat restrictions. Younger users cannot use chat, covering roughly one-third of the base.

(2) Gameplay limits. Age verification is required, with unverified accounts restricted to 'Minimal/Mild' titles and no communications, and the platform created dedicated kids accounts.

(3) Recommendation shift. The engine moved from 7-day monetization optimization to 28-day retention, reallocating exposure from high-monetization titles to high-retention, lower-monetization ones. In short, pay-heavy games were de-emphasized while large-DAU titles gained exposure.

On the above, especially the third change, we will keep watching. In theory this should favor healthier long-term growth, but the near-term revenue hit is material, and it is unclear whether de-emphasizing monetization will also sideline high-quality, monetization-heavy indie games. Tracking D28 retention over time should reveal the true effect.

2) Results kept sliding; FY guide withdrawn. Commercial headwinds became more visible in Q2 and, against a tough base, management guided Q3 bookings growth to turn negative.

At the same time, Roblox is lifting developer rev-share (and genres favored by adults, such as shooters, typically carry higher splits than hyper-casual). It is also investing in tech and new features to enhance UX (e.g., Roblox Reality). With revenue under pressure, profits will compress even faster.

Given the magnitude of change this year, management chose to sacrifice near-term monetization and scrapped the full-year outlook. For investors, expectations have deteriorated and visibility has decreased. Even the near-term bottom in results is hard to pin down for now.

3) Buyback underway, but not enough to draw capital. In May, management announced up to $3bn in repurchases, with ~$1bn over 12 months (implied 3% yield). In Q2 it bought back 8.2mn shares for about $380mn, a faster pace that signals a willingness to backstop the stock.

COIN 2Q26 First Take: Crypto stayed under pressure in Q2 on higher-for-longer rate expectations, geopolitical frictions, and Strategy’s selling, with a sharp pullback in Jun. In this backdrop, results are unlikely to be strong. Some sentiment-driven de-rating likely occurred ahead of the print.

Beyond macro, the slow progress on the CLARIFY Act also weighed on sentiment. It remains stuck awaiting a Senate floor vote. With the Aug recess and midterms ahead, passage this year looks unlikely, removing a key catalyst for Coinbase.

1) Trading kept sliding: trading revenue fell 19% YoY. The company did not disclose legacy trading volume metrics this quarter, but Coinbase spot volumes declined sharply YoY and QoQ, in line with the broader market. That said, management’s disclosed market share metric continued to rise.

2) Subscriptions also lagged, with soft guidance: non-trading subscription revenue is driven by stablecoin income, staking income, Base chain revenue, and interest income. Q2 came in at $560 mn (-12% YoY, -5% QoQ). The mix is smoother than trading but still weakened.

By line item, stablecoin income largely tracked USDC market cap. Staking was hit by a double whammy of lower token market caps and yields, posting a steep decline. Only custody fees and Base chain revenue were relatively stable (+4% QoQ, -5% YoY).

For Q3, management guides subscription revenue to be flat QoQ and below Street expectations. This implies a slower near-term recovery in USDC expansion as well as in ETH/SOL prices and staking yields.

3) Continued increase in stablecoin holdings: the stablecoin market also shrank in Q2 amid the crypto slump and the CLARIFY overhang. USDC market cap fell to $73.5 bn (-5% QoQ), with more pronounced swings than USDT. This likely reflects USDT’s broader use cases, including commerce payments, and a higher non-US user base, making it less sensitive to rates and CLARIFY timing.

Coinbase kept adding exposure, with average platform-held and custodied USDC at $20 bn, 26% of circulating supply, up from 25% in Q1. The company has renewed its partnership with Circle. Both sides remain tightly aligned as ecosystem partners.

4) Core profit turned negative: with revenue under pressure and relatively sticky opex, margins compressed. Despite cost controls, severance, consolidating Deribit, and new product investments drove R&D up 22%. As a result, gross profit less operating expenses turned negative in the quarter.

5) Buybacks slowed: Coinbase repurchased $81 mn in Q2 (810k shares at an avg. $148.94), largely during the Jun selloff. That compares with ~ $1.1 bn in Q1, effectively slamming the brakes. $Coinbase(COIN.US)