How Far Can Nasdaq Euphoria Run? Narrow Rally, Meta's Muse and Cheap Chips

It was a historic day for tech: the S&P 500 rose 1.49%, the Nasdaq 2.26%, and Bora Özkent's own portfolio climbed over 3%. Ten-year yields slipped, the dollar softened, Brent held just under $100 and Bitcoin pushed above $86,000 before easing to $85,000 yet stayed strong. Technically the tape strengthened too, with the S&P 500 and QQQ breaking above all moving averages and semis snapping a long slump, propelled by Meta's Muse momentum. Like a main boulevard lit up while side streets stay dark, the headline index glowed while breadth stayed cautious, but buyers were clearly back.

An AI-driven technical warning arrived at the same time: the advance was carried by very few names and breadth did not expand. Narrow leadership is usually read as fragility because healthy rallies want many stocks moving together. Özkent flips the script, arguing big rallies often start narrow, led by tech first, with money spreading later. He reads the market's response to the recent rate decision and Friday's late surge as constructive, and with geopolitical easing plus AI optimism the tilt stays upward, though no rally lasts forever. He stresses this is a forward view, not investment advice.

The record makes the picture more striking: while the Nasdaq hit an all-time high, 59.2% of S&P 500 constituents sit more than 20% below their own peaks, in bear-market territory, and 81.5% sit more than 10% below. Tech drags the index higher yet even inside tech names like Micron remain below prior highs. For Özkent that dispersion suggests the broad market is not expensive, like a shiny window display with discounted shelves in the back. The root cause is the economy: higher rates bite low-income consumers and housing, while high earners benefit from higher risk-free returns. Two economies at different speeds explain the gap inside the market.

Why rates bite less today is explained with a 2008 contrast. Then growth was housing-led and housing is rate-sensitive; now growth is AI-led and AI is far less rate-sensitive. AI infrastructure is largely funded from the companies' own free cash flow, with limited external debt, so higher rates have weak stopping power. Neo-cloud players like Oracle and CoreWeave are the more leveraged exceptions, but the core is self-funded. That is why AI spending looks set to continue, and Meta's Muse is presented as the freshest proof. From energy to chips to software, the spending loop seems to clear the rate wall for now.

At the center is Muse, Meta's consumer personal agent. Its U.S. store downloads outpaced ChatGPT's first mobile launch, not because the model is better but because distribution is. Codenamed Hatch inside the company, Muse reaches billions through Instagram, Facebook and WhatsApp and promises to solve everyday jobs. The example Özkent cites is telling: a user hands a hotel refund dispute to Muse, and the agent handles the correspondence, the complaint draft and the reimbursement end to end. How it works in plain steps: 1) understand the request, 2) reach email, calendar and payments from an isolated cloud computer, 3) spin up sub-agents to run the workflow, 4) ask for human approval and use a single-use card for any payment. Not the best language model, but as a product it spreads fast.

Behind the euphoria Meta's fundamentals still strain. To fund infrastructure the company has raised debt, free cash flow is under pressure and margins slipped last quarter. Yet the stock jumped 11.43% as investors read the new consumer agent as fresh proof of a growth story, tolerating near-term cash pressure for long-term platform optionality, like funding an expensive kitchen to lift a home's value. Özkent notes he has stayed bullish on Meta for a month and that the community benefited, with Zuckerberg's return to product focus underpinning the call.

The first trade war of the agent economy broke out the same day. Amazon said it would block Muse agents from shopping on its site on security grounds, a move Özkent calls defensive and short-sighted, a wall instead of adaptation. Shopify filled the gap, announcing it will let Muse agents shop across its merchant network. The imagined flow is simple: you spot an outfit on Instagram, tell the agent, it finds the item, matches the size and checks out with a single-use card upon your approval. Grokbot does similar jobs but at a smaller scale; X's reach is tiny next to Meta's trio, so scale, not just model quality, may decide who wins.

The biggest surprise was in the CPU link of the chip chain. Agent workloads like Muse consume far more CPUs than GPUs, with talk of a 40-to-1 CPU-to-GPU ratio, pushing Intel, AMD and designer ARM back into the spotlight. AMD touched a $1 trillion valuation during the day, Intel and ARM surged too. Özkent offers self-criticism here: he had called Meta's rise but missed its chain effect on CPU demand. Nvidia still gained about 2.5%, with infrastructure names up 5-10%. Eyes then turned to the Trump-Xi meeting, where Nvidia's H200 sale to China is a key item, with roughly 1.5 million units and $30 billion of incremental revenue in play; Washington has eased but Beijing's push for domestic chips keeps the outcome uncertain, a top catalyst for the week.

Physical limits of AI also featured. Many data-center projects are being paused or canceled, hurt by local political resistance and by energy bottlenecks. Two groups may benefit: nuclear and energy names on one side, former Bitcoin miners on the other. Hut 8 and Riot hold large power contracts and sites; Hut 8's lease to fully commercialize its Texas campus for AI, at about $9.8 billion reported by Reuters, shows the scale of the pivot. As Bitcoin hovers near $86,000 these miners must choose where to allocate capacity, mining or AI leasing; as energy grows scarce, harder Bitcoin production can support the price while power-rich miners find a new income lane.

Rockets face a similar scarcity story. SpaceX slowed Falcon 9 output to transition toward Starship, tightening launch supply; after a record 165 Falcon 9 flights in all of 2025 the company is at 103 so far this year and may end around 150-160. Demand from states and private satellites keeps rising and few vehicles match Falcon 9's reliability. Smaller launchers like Rocket Lab try to fill the gap, focused on lower orbits but still catching a scarcity premium with a sharp rise yesterday. Özkent argues that as geopolitical risks ease, money may flow faster into high-beta space names. A possible contact between Trump and Iran's president at the United Nations gathering in New York this week is being priced as one such easing.

Is the market expensive then gets a clear no from Özkent, with math. From the start of the year to September 21 forward earnings expectations rose 28.3% while the S&P price index rose only 13.4%, leaving an 11.6% gap that, historically, price tends to close by following earnings. Not every sector shares the earnings lift and war, oil, rates and policy noise still weigh, yet the broad tape does not look pricey. The cheapest pocket is chips: forward price-to-earnings around 16.5 times for semis versus 19.4 for the S&P average, so the fastest growers trade at a discount. That basket includes Micron, Nvidia, AMD, ARM and a recovering Intel plus optical names lower down. Cybersecurity stays expensive at the margin, but for semis and hardware Özkent sees room ahead, keeping the upward bias even if a pullback comes.