Leveraged NuScale ETFs Jump More Than 30% as the Nuclear Trade Turns Speculative

Apple approaches its November 2 fiscal fourth-quarter report with a familiar advantage and a newer problem. Product demand has been strong, but the cost of supplying it is rising just as a one-time tariff benefit begins to fade.

The June quarter set a company record. Apple reported revenue of $109.4 billion, up 16% from a year earlier, and diluted earnings of $2.02 a share, up 29%. Gross margin reached 50.1%, but about two percentage points came from tariff refunds. The same refunds added roughly $0.11 to earnings per share, according to Apple's release.

That makes the September-quarter comparison less straightforward than the growth rate suggests. Management guided revenue growth of 9% to 11% and gross margin of 47% to 48%, including about one percentage point of tariff benefit. Consensus figures cited by 24/7 Wall St. call for revenue of $113.36 billion and normalized earnings of about $1.98 a share, versus $102.47 billion and $1.85 a year earlier.

Memory costs are the main pressure point

Apple said higher memory prices accounted for more than all of the sequential change in adjusted gross margin on its July call. Advanced-chip and memory supply were also constraining Macs, iPads and some iPhone availability. Strong demand can support revenue, but it does not automatically protect profit if scarce components become more expensive.

The company guided to mid-teens iPhone growth for the September quarter. The newly launched $1,999 iPhone Duo sold out its initial online preorder allocation within minutes, according to the source report, creating a strong starting point for the holiday period. The important question is whether Apple can replenish supply without using discounts or absorbing another step-up in component costs.

Services provide a partial counterweight. Services revenue grew 12% to $30.7 billion in the June quarter, although mobile gaming slowed. Higher-margin services and iCloud subscriptions can offset some hardware pressure, but the full compute cost of Apple's new Siri AI features remains uncertain. Regulatory work has also delayed the rollout in the European Union and China, limiting how quickly the company can turn the product into a broader ecosystem benefit.

A premium valuation raises the standard

Apple was valued near $5.07 trillion in the source article, after a 26.65% gain in 2026. At about 35 times forward earnings, investors are already paying for durable demand, controlled costs and a smooth leadership transition to CEO John Ternus.

A revenue or earnings beat alone may not settle the debate. Apple has exceeded earnings estimates for ten straight quarters, yet 24/7 Wall St. calculated that the shares fell an average of 1.16% on the following day. That pattern shows why the market will focus on what happens next rather than only on the reported quarter.

Three disclosures would change the investment picture most: December-quarter gross-margin guidance, the scale and duration of memory and advanced-node constraints, and evidence that Siri AI can deepen services revenue without driving a disproportionate increase in infrastructure spending. If margins hold while Apple meets holiday demand, the premium valuation has operating support. If another cost warning follows the fading tariff benefit, even strong iPhone sales may not be enough.

Intel shares fell 10% through Thursday of last week while the Nasdaq 100 lost about 3%. The move looked like a verdict on Intel, but most of the immediate pressure came from a semiconductor-wide selloff after Lam Research warned that supply constraints limited its December-quarter outlook.

Lam still reported record September-quarter revenue of $8.22 billion and guided to $8.6 billion at the midpoint for the next quarter. The combination matters: demand remained strong, yet investors punished suppliers that could not convert all of it into near-term revenue. Intel faces the same distinction between demand and deliverable capacity.

Intel's operating backdrop is stronger than the share move alone suggests. The company reported second-quarter revenue of $16.1 billion, up 25% from a year earlier, and guided third-quarter revenue to $15.8 billion to $16.8 billion. Intel Foundry revenue rose 31% to $5.8 billion, although most of that business still came from Intel's own product groups. External foundry revenue remained small.

Packaging is the nearer commercial test

Intel expects advanced packaging to become one of the first meaningful external foundry businesses. Its EMIB-T technology connects multiple chip components in one package, a capability that is increasingly important for large AI systems. Management has indicated that related revenue should improve in the second half of 2027 and has discussed gross margins near 40%.

The risk is that customers will not wait. On Taiwan Semiconductor Manufacturing's October 15 earnings call, CEO C.C. Wei was asked whether TSMC could support customers that run short of Intel's EMIB-T capacity. Wei emphasized TSMC's own advanced-packaging scale and yields. Intel has not publicly named a major EMIB-T customer, so evidence of a binding external order would matter more than another general statement about demand.

Intel also disclosed in its second-quarter filing that demand for data-center products exceeded available supply and that industry-wide constraints could persist into next year. That is supportive for pricing, but only if the company can add capacity without sacrificing yields or margins. A fatal October 14 explosion at Intel's Kiryat Gat site added an operational concern; as of the source article, Intel had not disclosed whether production was affected.

What the next report must establish

Intel reports third-quarter results on October 29. Analysts cited by TIKR expect roughly $16.4 billion of revenue, within the company's guidance range. The more revealing figures will be fourth-quarter gross-margin guidance, progress on advanced-packaging capacity and any evidence of external foundry customers.

TIKR put Intel at about 51 times forward earnings after the decline, close to its five-year average, with a $117 analyst consensus target. Those figures are estimates, not proof that the shares are cheap. Intel's stock had still gained 136% in 2026 by the source's measurement, leaving substantial turnaround expectations in the price.

The selloff therefore creates a clearer test rather than an automatic buying signal. If Intel names a packaging customer and holds gross-margin guidance near 42%, the market may view the capacity constraint as temporary. If TSMC captures customers that Intel cannot serve, packaging could become another delayed foundry promise.

Wells Fargo, Arm Holdings and Intuit are among six more companies that have blocked OKXICE from offering tokenized versions of their shares, bringing the disclosed objection count to 22. Reuters reported the additions on October 23 after reviewing a regulatory document. The companies did not explain their decisions, so the objections should not be read as a verdict on every form of stock tokenization.

The immediate consequence is narrower than the headlines suggest. The objections apply to OKXICE, the planned venue jointly owned by Intercontinental Exchange and crypto exchange OKX. They do not prevent the same issuers from supporting another tokenization structure later. Still, 22 objections against a proposed list of 63 stocks show that issuer consent can materially shrink a venue before trading begins.

How the SEC framework gives issuers control

The Securities and Exchange Commission approved a five-year innovation exemption on September 17 for qualifying tokenized securities venues. The exemption permits limited trading of tokenized U.S. listed stocks in a permissioned environment while the SEC studies the market structure. Under the process described in the OKXICE disclosures, a venue must notify an issuer before listing a third-party token tied to its shares and wait at least 30 calendar days. A timely written objection stops that token from trading on the venue.

That safeguard has produced a rapid series of opt-outs. OKXICE disclosed nine objectors on October 15, then 16 on October 19. The earlier list included Broadcom, Verizon, UnitedHealth, Microsoft, Alphabet and Goldman Sachs. The latest Reuters report added Wells Fargo, Arm and Intuit among six more companies, although the other three were not named publicly in the reporting reviewed for this article.

The venue's proposed design also matters. OKXICE plans to use permissioned Uniswap v4 pools on XLayer, with stock tokens paired against stablecoins. Only wallets holding the venue's nontransferable access token would be eligible to trade. The venue has not announced a launch date, and its disclosures state that it is not registered with the SEC as a conventional exchange for these activities.

What investors can conclude

Tokenization advocates argue that blockchain-based shares could trade around the clock and settle faster. The objections do not disprove those benefits. They do show that technological permission and regulatory permission are not the same as commercial acceptance.

For Intercontinental Exchange, the development is strategically relevant but not yet financially measurable. The venture can still launch with issuers that do not object, including several large technology companies listed in the October 19 notice. The unresolved question is whether a smaller roster attracts enough liquidity to make 24-hour trading useful. Until OKXICE names a launch date, publishes an updated issuer list and shows trading economics, the objection count is evidence of adoption friction rather than a quantified earnings risk.

Arista Networks' AI networking story is no longer supported only by forecasts. Second-quarter 2026 revenue reached $3.036 billion, up 37.7% from a year earlier, while non-GAAP operating margin increased to 49.9% from 48.8%. The company also guided to about $3.3 billion of third-quarter revenue and non-GAAP earnings of $1.06 to $1.08 a share.

The strategic bet is that Ethernet can take a larger role inside AI clusters that have often relied on Nvidia's InfiniBand. Arista combines merchant silicon with its EOS operating system and Etherlink networking products, giving customers an open fabric that can connect accelerators from multiple vendors. Management says Etherlink has passed 100 cumulative customers, compared with only four or five in 2024.

Arista's latest product cycle supports that argument. The company introduced 1.6-terabit AI fabric platforms for scale-up, scale-out and scale-across networks, including liquid-cooled systems. It says the new 7060XE7 series can provide as much as 100 terabits per second of system bandwidth and use linear pluggable optics to cut interconnect power by about 60% compared with traditional pluggable optics.

Demand visibility is improving, but the balance sheet reveals how much execution is required. At June 30, Arista held $2.29 billion in cash and $11.05 billion in marketable securities. Deferred revenue was $6.87 billion, and total remaining performance obligations were $8.4 billion, with about 91% expected to become revenue within two years.

The same filing showed $9.7 billion of non-cancellable purchase commitments, including $9.4 billion due for receipt within 12 months. Those commitments can protect supply during tight memory and silicon markets, but they also expose Arista if customers delay AI projects or demand shifts faster than inventory can be adjusted. The company explicitly warns that customer orders may not match its forecasts.

Customer concentration adds another risk. Two customers each accounted for more than 10% of annual revenue in each of the last three years. The larger represented 26% of 2025 revenue and the other 16%. Arista also depends on a predominant merchant-silicon vendor, so a supply disruption or a change in a large customer's architecture could affect growth and margins at the same time.

This makes the investment case more balanced than the original claim that Arista is simply breaking an Nvidia monopoly. Ethernet adoption is broadening, and Arista is converting that demand into unusually strong revenue growth and operating profitability. Yet the $9.7 billion supply commitment means management is placing a large, binding bet on forecasts that customers can still revise.

The next useful evidence is not another customer-count headline. Investors should compare third-quarter revenue and margin results with guidance, then watch whether deferred revenue converts without a buildup of excess inventory. If Ethernet keeps gaining share while purchase commitments remain aligned with orders, Arista can justify more of its premium valuation.

Honeywell Technologies has put a measurable target on its first Forge Plant Cognition deployment: more than $60 million in estimated annual gross operating benefits at Dangote's Lekki refinery when the plant runs at full capacity. The figure gives investors a concrete way to judge the industrial AI launch, but it is a customer-benefit estimate, not Honeywell revenue.

The platform combines the refinery's live operating data with Honeywell UOP's process expertise, engineering calculations and physics-based models. NVIDIA supplies AI models and simulation tools, including Nemotron, Omniverse and Cosmos, while Dell Technologies provides the computing infrastructure. Honeywell says the system can identify bottlenecks, flag emerging equipment problems and recommend operating changes to improve throughput, yield and reliability.

The promised economics need careful framing. Honeywell's preliminary estimate is more than 29 cents of gross operating benefit per barrel. The company says the calculation is specific to the Lekki facility, excludes implementation costs and is not guaranteed. Actual results will depend on utilization, crude mix, product margins, baseline performance, data quality and Dangote's execution. No contract value or expected revenue contribution for Honeywell was disclosed.

That distinction matters because the deployment is an operating test, not yet proof of a scalable software business. A successful installation could give Honeywell a reference site for selling the platform to other refineries, petrochemical plants and liquefied natural gas operators. Weak or hard-to-measure results would make the $60 million headline much less useful as evidence of customer return on investment.

Dangote currently lists the refinery at 650,000 barrels a day and plans to expand capacity to 1.4 million barrels a day within three years with Honeywell's help. Forge Plant Cognition is intended to support the existing plant while that expansion proceeds. The companies also expect the tools to transfer specialized process knowledge to local engineers as Dangote works toward a workforce that is 70% Nigerian.

For Honeywell shareholders, the immediate catalyst is modest because pricing, margins and deployment timing remain undisclosed. The more important signal is that the company is tying its Forge strategy to a large, operating industrial asset and allowing the result to be judged against an explicit per-barrel target. The next evidence should be realized savings, uptime or throughput improvements, followed by details on how those gains translate into recurring revenue for Honeywell.