To The Moon

The market: AI shares rose first, then earnings moved the question from demand to returns

The $Invesco QQQ(QQQ)$ gained 1.05 per cent to US$725.20 on Wednesday, while the $S&P 500(.SPX)$ rose 0.58 per cent to 7,757.17. $SpaceX(SPCX)$ gained 7.79 per cent, $SUPER MICRO COMPUTER INC(SMCI)$ 9.55 per cent, N $NVIDIA(NVDA)$ 2.54 per cent and $Advanced Micro Devices(AMD)$ 3.36 per cent. These were regular-session moves before Meta and Tesla reported. Investors were initially trading the combined message from SpaceX's results, Supermicro's higher guidance and GTC Berlin: demand for AI infrastructure has not disappeared.

After Meta and Tesla delivered their numbers, the question advanced. It is no longer simply whether demand exists. It is which business model can turn the same AI spending — through advertising, cars, servers and networks — into free cash flow first. Both reports offered support, and both left a gap.

Meta: advertising is working, but the data-centre bill exceeded quarterly free cash flow

$Meta Platforms, Inc.(META)$ rose 0.47 per cent to US$697.52 in the regular session. Third-quarter revenue was US$64.601 billion, up 26 per cent year on year and above market expectations. Ad impressions increased 10 per cent and average price per ad rose 14 per cent. The company guided fourth-quarter revenue to US$73 billion to US$76 billion, with the midpoint slightly above the previous analyst consensus of roughly US$73.6 billion. The advertising engine did not slow under the weight of AI investment, while products including Muse are widening the route to users.

Cash flow tells a different story. Operating cash flow was US$32.79 billion, while capital expenditure reached US$34.91 billion, leaving free cash flow negative at US$2.12 billion. Meta also booked a US$9.89 billion litigation charge, and operating margin fell to 20 per cent from 40 per cent a year earlier. The company announced a new class of non-voting Class C common stock as well. AI is helping revenue grow, but the quarter did not show that this growth can yet fund the infrastructure bill consistently.

Tesla: profit weakened, but cash flow held

$Tesla Motors(TSLA)$ gained 3.41 per cent to US$386.47 before its report. Third-quarter revenue was US$28.05 billion, broadly flat year on year, and adjusted earnings were US$0.48 per share. GAAP operating income was US$0.7 billion and net income was US$0.8 billion, while core automotive revenue declined 3 per cent. Spending on AI, robotics and new products continued to weigh on profit, so the income statement was not an easy one.

Cash generation was firmer: operating cash flow reached US$6.3 billion, free cash flow was US$1.5 billion and cash and investments increased by US$1.9 billion. Tesla also said FSD's attach rate exceeded 75 per cent of new North American deliveries, Cybercabs joined the Austin Robotaxi fleet on 3 September, and cumulative unsupervised Robotaxi miles increased eightfold quarter on quarter. That provides an observable path from AI spending to product use. The unresolved issue is whether usage can become a profit pool large enough to offset research and capital expenditure while automotive revenue remains under pressure.

Nvidia: this was not another chip launch; it was an attempt to secure the factories

$NVIDIA(NVDA)$ rose 2.54 per cent to US$221.97. At GTC Berlin, Jensen Huang described the AI economy as five layers — energy, computing, AI factories, models and applications — and framed telecommunications networks as the next distributed AI infrastructure. The more concrete move came in manufacturing: Nvidia announced capacity expansion with Amkor, Ibiden, KYEC, Samsung Electro-Mechanics, SPIL and Unimicron across substrates, advanced packaging, testing and optical engines.

The commitments are measurable. Aggregate substrate capacity is planned to rise to 2.5 times its current level over three years. Amkor aims to expand advanced-packaging capacity by more than 300 per cent by the end of 2028. Nvidia is paying KYEC a US$1.1 billion capacity-reservation advance; this is Nvidia's cash outflow, not a cap on KYEC's total future revenue. SPIL targets a fifteenfold increase in optical-engine capacity for co-packaged optics by mid-2027. Bulls see order visibility. The risk is that capacity secured in advance can amplify a cycle if customer free cash flow fails to keep pace.

Supermicro: demand was validated, but the margin remains in the low teens

$SUPER MICRO COMPUTER INC(SMCI)$ gained 9.55 per cent to US$35.57. The company raised its preliminary first-quarter financial year 2027 revenue estimate to US$18.5 billion to US$19.5 billion from US$14.5 billion to US$15.5 billion, and lifted estimated gross margin to 12.0 per cent to 12.5 per cent from 10.4 per cent to 10.8 per cent. It attributed the changes to higher AI-platform demand and customer and product mix. Raising revenue and margin together shows that server deployment is more than a chipmaker's narrative.

The figures remain preliminary and unaudited rather than signed orders, and the company did not say how much revenue would flow to Nvidia or AMD; formal results are due on 3 November. Even at the revised level, gross margin is only around 12 per cent. Supermicro therefore looks more like a high-throughput infrastructure contractor than a platform with chip-like pricing power. Strong demand can lift revenue, but execution, financing and component costs determine how much profit stays behind.

The next test: not who is still buying, but who can pay from internal cash

$Western Digital(WDC)$ reports after the close on Thursday, testing whether storage demand is spreading beyond AI servers into the broader enterprise market. Nebius will give investors another opportunity to compare neocloud revenue growth with capital expenditure and funding costs. When Alphabet, Microsoft and Amazon report next week, the key measure will not be cloud growth alone, but the distance between capital expenditure and free cash flow.

Taken together, today's evidence does not say AI demand is weakening. It says demand is leaving very different cash outcomes across business models: Meta's advertising revenue is strong but free cash flow is negative; Tesla's profit is under pressure but cash flow is positive; Nvidia is committing money to supply-chain capacity in advance; and Supermicro is absorbing deployment at a relatively thin margin. The next stage of the trade depends on finding which layer can compound investment without requiring the next round of external funding.

This is personal analysis and not investment advice.

💬 【Talking Point】

If AI demand stays strong but free cash flow does not follow, what breaks first: valuations, capital-expenditure plans or supplier margins?

💰 【Bounty】

Drop your view and reasoning in the comments and there are Tiger Coins in it for you! 🎁

🔔 Better shared than saved — tag a friend and split the coins!