Musk Clears FTC Review to Buy Mesh Optical, Grabbing the AI Data Center's Quietest Bottleneck

An early-termination notice on June 25th hands Musk the go-ahead to acquire a former-SpaceX-engineer startup building optical transceivers, even as X Money reaches its first paying users and Starlink weighs a direct retail phone service.

Everyone fixates on chips and power when they argue about who wins AI. Almost nobody talks about the wiring between the racks. That is exactly where Musk just spent money. Regulators cleared him to acquire Mesh Optical Technologies, a startup building optical transceivers that move data with light instead of copper inside high-performance computing clusters. On its own it looks like a small, technical tuck-in. Read against the rest of the week, it looks like a man methodically buying the parts of the AI stack that consensus keeps ignoring.

Key Takeaways

• Mesh Optical Technologies cleared its antitrust waiting period on June 25th via an early-termination notice, giving Musk a green light to acquire a startup founded by ex-SpaceX engineers.
• Backed by a Series A of more than 50 million dollars led by Thrive Capital, Mesh builds optical links aimed at lower latency and better power efficiency inside data centers.
• Nobody has said which entity ends up owning it. xAI, SpaceX, Tesla, or a separate Musk vehicle are all live options, and the filing discloses no valuation or closing date.
• X Money went live to a slice of US Premium Plus subscribers, with deposits insured up to 250,000 dollars through Cross River Bank, a Visa debit card tied to your handle, 3 percent cashback, and no foreign transaction fees.
• Starlink is weighing a retail mobile service sold straight to US consumers, a move that would put it head-to-head with Verizon, AT&T, and T-Mobile instead of partnering with them.
• Index mechanics are about to hit: SpaceX joins the Nasdaq 100 before the open on July 7th, and FTSE Russell inclusion implies roughly 3 billion dollars in passive demand.
• Tesla's own compiled consensus points to 406,000 vehicle deliveries and 13.8 gigawatt-hours of storage for Q2 2026, with Goldman Sachs already modeling 420,000 cars.
• A Tesla Semi that stickers around 290,000 dollars could drop to as low as 50,000 dollars for California small fleets stacking state voucher and clean-fuel programs.

---

Why Optical Wiring Is the Constraint Nobody Prices

An AI cluster is bottlenecked by four things: the chips, the power feeding them, the cooling pulling heat away, and the networking stitching thousands of accelerators into one machine. The first three get all the airtime. The fourth is where Mesh lives. Optical transceivers convert electrical signals into light so data can cross a data center faster and with less energy burned per bit. As models scale and clusters push past tens of thousands of GPUs, the links between nodes stop being plumbing and start being a ceiling on how big and how efficient your training run can be.

That is the whole reason this deal is interesting. Musk is not buying a flashy consumer product. He is buying a layer of infrastructure that becomes a hard constraint precisely when everyone else is still fighting over silicon.

The Mellanox Comparison Is the Right Lens

The cleanest analogy is NVIDIA swallowing Mellanox. NVIDIA didn't buy Mellanox for glamour. It bought the interconnect because owning the networking around the chips is how you control the performance of the whole system. Applied here, pulling Mesh in-house could give a Musk company a US supply chain for optical transceivers at a moment when a lot of the competing components are manufactured in China.

I think that supply-chain angle is underrated. If you believe AI infrastructure becomes strategically contested the way semiconductors already have, then owning a domestic source of a critical link is worth more than the modest price tag suggests. It is a vertical-integration bet on a component most investors can't even name.

The Unanswered Question: Who Actually Owns It

Here is what the filing does not tell us, and it matters more than the clearance itself. We don't know the valuation. We don't know the closing timing. And we don't know which Musk entity ends up holding the asset. Mesh is early. It still has to prove its technology works at hyperscale, not just in a lab or a Series A pitch.

The assignment is the real signal. If Mesh gets folded into xAI or SpaceX, that tells you the intent is a specific AI build-out, not a passive stash. And if it lands inside xAI, it reinforces why I think that company is structurally better positioned than the market believes. Owning the bottleneck layer, rather than renting it, is the kind of advantage that compounds quietly for years before anyone notices.

X Money Finally Ships, and the Metric That Matters

X started rolling X Money out to a subset of US Premium Plus subscribers, its first broad consumer launch after months of internal testing. The feature set is a real bank-adjacent product: peer-to-peer payments, bill pay, wires, mailed checks, deposits insured up to 250,000 dollars through Cross River Bank, a Visa debit card tied to your X handle, 3 percent cashback, and zero foreign transaction fees. This is the everything-app thesis made concrete, and it is central to X's plan to earn money beyond ads and subscriptions.

But I'd ignore the headline rollout numbers. Payments are a different animal from running a feed. You inherit fraud, compliance, state-by-state licensing, and customer-support obligations the instant you touch someone's money. The number I'd actually watch is repeat transaction behavior. Does moving money inside the social graph become a native habit, or a one-time novelty? Everything hinges on that, and it arrives roughly two months after the April target Musk floated for early access.

Starlink Wants to Sell You a Phone Plan

The bigger swing this week is telecom. SpaceX is reportedly considering a Starlink-branded retail mobile service sold directly to US consumers, an idea floated to investors during its recent IPO roadshow. Today Starlink reaches phones in the US through a partnership with T-Mobile. A direct retail product changes the posture entirely, putting Starlink into the ring with Verizon, AT&T, and T-Mobile rather than riding on one of them.

The logic is control. Starlink already owns the satellite infrastructure, holds spectrum-related assets, and carries a brand built on covering the places terrestrial networks forgot. A direct service captures the customer relationship and expands the market well beyond fixed broadband. The catch is enormous: a terrestrial build-out means regulatory approval, real capital, network planning, and untangling existing carrier relationships, all against incumbents with dense urban networks and billing systems refined over decades. This is the kind of large-market disruption investors underappreciate early precisely because the hard part looks impossible right up until it doesn't.

Index Inclusion Is Mechanics, Not a Thesis

SpaceX joins the Nasdaq 100 before trading opens on July 7th, with tracking funds expected to start buying after the July 6th close. FTSE Russell is adding it to its US indexes as well, implying an estimated 3 billion dollars in passive demand, and some estimates put Nasdaq-linked buying at 8 to 10 billion. With a limited float after the mid-June IPO, that forced demand can tighten supply and juice near-term volatility.

I want to be clear about what this is. It is a mechanical event, not a change in the business. Index committees moving a name in or out doesn't make it a better or worse company. The passive flow is a short-term trading factor that fades once the rebalance window closes. The thing that actually matters is whether SpaceX keeps validating the operating thesis: launch cadence, Starlink growth, Starship execution, and margins. SpaceX also remains outside the S&P 500 while it works through eligibility timing.

Tesla's Q2 Baseline: Watch the Gigawatt-Hours

Tesla's company-compiled consensus sets the Q2 2026 bar at 406,000 vehicle deliveries and 13.8 gigawatt-hours of energy storage deployed. Goldman Sachs has already nudged its delivery forecast up to 420,000 from 405,000, citing regional data and possible outperformance. The delivery figure will dominate the reaction, and 406,000 would show sequential improvement without, on its own, silencing worries about pricing pressure, competition, and regional demand.

The storage number is the one I think the market keeps mispricing. 13.8 gigawatt-hours is the quiet detail that improves gross margin mix, and it points at a future where AI-driven electricity demand turns Tesla Energy into a far larger earnings driver than the car business gets credit for. Elsewhere in the portfolio the same integration story keeps showing up: a Tesla Semi that lists near 290,000 dollars can fall to as low as 50,000 for California small fleets stacking incentives, SpaceX ran a 15-second single-engine static fire on Starship Ship 40 ahead of its thirteenth test flight, and xAI pushed Grok into the T3 coding tool without a separate API key. Different companies, one pattern. Own the layer everyone else rents.