September was supposed to be a bad month for richly valued tech. The Fed hiked, the 10-year Treasury yield pushed above 5%, and oil ran past $100. Yet the biggest AI companies mostly held up, and some had their best stretch in years. The market is no longer buying "AI" as one theme. It is picking winners company by company, rewarding visible revenue and punishing spending that has not yet paid off. Here is where the leaders stand heading into October.
Meta: from spender to story
Meta has been the standout of the month. The launch of Muse, its AI assistant that can book tickets, shop and send messages on a user's behalf, went straight to the top of Apple's App Store, with millions of downloads in under two weeks. Analysts responded with price-target upgrades, and the shares are on course for one of their strongest months in years.
What changed is the narrative. For most of the last two years, investors saw Meta's AI budget, guided at roughly $115-135 billion of capital spending this year, as a cost with no obvious product attached. Muse gives the market something it can measure: users, engagement and, through paid tiers, a direct revenue line. That is exactly the kind of proof the market is now demanding from every AI spender.
Oracle: the backlog trade
Oracle reported on 10 September, and the headline number was its contracted backlog. Remaining performance obligations, meaning revenue customers have already signed up for but Oracle has not yet delivered, rose to about $664 billion, after more than $30 billion of new AI cloud contracts in the quarter. Total revenue grew about 30% to $19.3 billion, and cloud infrastructure revenue more than doubled.
The catch is timing. Oracle expects only around 13% of that backlog to turn into revenue over the next twelve months, with most of it landing two to five years out. A large share is tied to OpenAI's Stargate data-center buildout. So Oracle is effectively a leveraged bet that its biggest customers keep paying, and it is funding the buildout with heavy borrowing. The stock tends to trade on every headline about OpenAI's finances.
Nvidia: the $5 trillion benchmark
Nvidia remains the largest company in the world, with a market value above $5 trillion. Its late-August results showed revenue of $96.2 billion for the quarter, more than double a year earlier, with data-center sales making up nearly all of it. It guided higher again for the current quarter.
The debate is no longer about demand. It is about margins and about how long its customers can keep spending at this pace. The company expects gross margins to dip into the low 70s later this fiscal year, partly because memory costs are rising. With the stock priced for continued perfection, even good news can produce muted reactions. That makes Nvidia's own earnings less of a catalyst than its customers' capex guidance.
Microsoft and Alphabet: steady, but watched on spending
Microsoft shares have climbed roughly 30% over the past three months, helped by strong uptake of its AI app-building platform and new infrastructure commitments abroad. Alphabet has had a choppier run. It came into September after its longest monthly losing streak in over a decade, then recovered on fast Google Cloud growth and favourable ad-tech court rulings, only to stall near $350 as investors weighed a capex bill of around $180 billion this year against rapid-fire Gemini model releases.
The private giants waiting in the wings
Two of the most valuable AI companies are still private. Anthropic, the developer of the Claude models, has filed for an IPO reportedly targeted for October. OpenAI, which also filed in June, is now said to be leaning toward a 2027 listing. Listings of this size would give public investors direct access to model developers for the first time, and could pull money away from the existing AI names while the deals are being marketed.
What this means for traders
The bottom line
The AI trade has matured. The market is still willing to pay up, but it now wants evidence: products people use, contracts that turn into revenue, and margins that hold. September's winners showed that evidence. Heading into October's earnings season, the question for every AI company is simple: where is the return on all that spending?
Meta: from spender to story
Meta has been the standout of the month. The launch of Muse, its AI assistant that can book tickets, shop and send messages on a user's behalf, went straight to the top of Apple's App Store, with millions of downloads in under two weeks. Analysts responded with price-target upgrades, and the shares are on course for one of their strongest months in years.
What changed is the narrative. For most of the last two years, investors saw Meta's AI budget, guided at roughly $115-135 billion of capital spending this year, as a cost with no obvious product attached. Muse gives the market something it can measure: users, engagement and, through paid tiers, a direct revenue line. That is exactly the kind of proof the market is now demanding from every AI spender.
Oracle: the backlog trade
Oracle reported on 10 September, and the headline number was its contracted backlog. Remaining performance obligations, meaning revenue customers have already signed up for but Oracle has not yet delivered, rose to about $664 billion, after more than $30 billion of new AI cloud contracts in the quarter. Total revenue grew about 30% to $19.3 billion, and cloud infrastructure revenue more than doubled.
The catch is timing. Oracle expects only around 13% of that backlog to turn into revenue over the next twelve months, with most of it landing two to five years out. A large share is tied to OpenAI's Stargate data-center buildout. So Oracle is effectively a leveraged bet that its biggest customers keep paying, and it is funding the buildout with heavy borrowing. The stock tends to trade on every headline about OpenAI's finances.
Nvidia: the $5 trillion benchmark
Nvidia remains the largest company in the world, with a market value above $5 trillion. Its late-August results showed revenue of $96.2 billion for the quarter, more than double a year earlier, with data-center sales making up nearly all of it. It guided higher again for the current quarter.
The debate is no longer about demand. It is about margins and about how long its customers can keep spending at this pace. The company expects gross margins to dip into the low 70s later this fiscal year, partly because memory costs are rising. With the stock priced for continued perfection, even good news can produce muted reactions. That makes Nvidia's own earnings less of a catalyst than its customers' capex guidance.
Microsoft and Alphabet: steady, but watched on spending
Microsoft shares have climbed roughly 30% over the past three months, helped by strong uptake of its AI app-building platform and new infrastructure commitments abroad. Alphabet has had a choppier run. It came into September after its longest monthly losing streak in over a decade, then recovered on fast Google Cloud growth and favourable ad-tech court rulings, only to stall near $350 as investors weighed a capex bill of around $180 billion this year against rapid-fire Gemini model releases.
The private giants waiting in the wings
Two of the most valuable AI companies are still private. Anthropic, the developer of the Claude models, has filed for an IPO reportedly targeted for October. OpenAI, which also filed in June, is now said to be leaning toward a 2027 listing. Listings of this size would give public investors direct access to model developers for the first time, and could pull money away from the existing AI names while the deals are being marketed.
What this means for traders
- Dispersion is the trade. AI stocks no longer move as a block. Companies showing products and revenue (Meta's Muse, Oracle's backlog) are being rewarded, while companies showing only spending are being questioned. Pair and relative-value ideas work better than betting on the theme as a whole.
- Capex guidance is the key catalyst. The big five cloud and AI spenders are guiding to well over $600 billion of combined capital spending this year. Any hint of a cut would hit chipmakers and suppliers first. Any increase would test investors' patience with the spenders.
- Watch yields. High-growth stocks are long-duration assets, meaning most of their value sits in profits far in the future. When real yields rise, those future profits are worth less today. September showed AI leaders can withstand 5% yields when they have a clear story, but those without one are exposed.
- Headline risk is concentrated. So much of the ecosystem depends on a few buyers (the hyperscalers) and a few customers (OpenAI and other model labs) that one financing or contract headline can move a dozen stocks at once.
The bottom line
The AI trade has matured. The market is still willing to pay up, but it now wants evidence: products people use, contracts that turn into revenue, and margins that hold. September's winners showed that evidence. Heading into October's earnings season, the question for every AI company is simple: where is the return on all that spending?
clean
by ai-agent
— Support AI-stocks news pairs, user-directed 2026-09-25