A sharp selloff in software shares alongside renewed strength in chipmakers and data-center infrastructure names is reinforcing one of 2026’s defining market trades, as investors increasingly rotate toward companies monetising the AI boom now and away from those seen as vulnerable to disruption by the same technology.

A powerful divergence has once again opened up across the technology sector, with hardware and AI infrastructure stocks attracting fresh institutional money while software names continue to come under sustained pressure.

The split, highlighted by CNBC’s Jim Cramer on Thursday, has rapidly re-emerged as one of Wall Street’s defining narratives after briefly taking a back seat during the Iran conflict and the ensuing market volatility.

In Cramer’s words, “I’m talking about the enterprise software empire that’s being toppled by hardware stocks and AI,” adding that, “this war in tech, more than the actual war in Iran, has captivated Wall Street.”

Thursday’s price action made the split impossible to ignore. Software stocks absorbed another beating. Salesforce fell nearly 3 percent, Adobe dropped close to 4 percent, and the iShares Expanded Tech-Software Sector ETF (IGV), the benchmark many big institutions use to express views on the sector, plunged more than 4 percent.

Cybersecurity heavyweight CrowdStrike got dragged down 7.5 percent simply because it sits inside the ETF, even though its business is more defensive than pure SaaS.

Cramer noted the IGV’s sharp move as a clear sentiment gauge: “This ETF is the primary way that big institutions bet on or bet against software.”

On the other side of the ledger, the winners were the companies supplying the concrete, chips, optics, and networking gear needed to power massive AI data centers. Marvell Technology and Intel each rose nearly 5 percent. Corning, a key provider of specialty materials for data center infrastructure, gained 2.85 percent.

“If you’re in the software camp, you’re being treated as if you’re ready for the embalmer,” Cramer said with characteristic flair. “If you are in the hardware and AI camp, you’re headed for the pantheon of greatness.”

The divergence isn’t a one-day anomaly. The IGV has suffered one of its worst stretches in years, plunging more than 24 percent in the first quarter of 2026 — its steepest quarterly drop since the 2008 financial crisis. Broader software indices have seen hundreds of billions wiped out since early February, when Anthropic’s rollout of Claude Cowork and its industry-specific plugins ignited what traders dubbed the “SaaSpocalypse.”

Investors suddenly feared that sophisticated AI agents could automate complex workflows in sales, legal, finance, and data analysis, potentially eroding the need for expensive per-seat software licenses and professional services.

That fear has lingered. Even as some software executives argue the panic is overdone and that AI will ultimately enhance rather than replace their platforms, the market has repriced growth expectations aggressively. Multiples have compressed sharply for names like Salesforce, Adobe, Workday, and ServiceNow, with several down 25-40 percent year-to-date at points this year.

By contrast, the hardware side continues to ride the wave of exploding capital expenditure on AI infrastructure. Gartner and other forecasters see spending on AI-related data centers, networking, power, and chips potentially reaching well over a trillion dollars in 2026. Companies like Marvell have posted strong data-center revenue growth and secured high-profile partnerships, including ties to Nvidia’s ecosystem, that position them to capture share in custom silicon and high-speed interconnects.

Cramer suggested this bifurcation has staying power, at least in the near term.

“Here’s the bottom line: maybe tomorrow we’ll return to the worldwide narrative, whether it’s war or peace in the Middle East,” he said. “But, for now, it’s just another day when hardware slew software like Cain slew Abel and all I can do is say get used to it.”

The takeaway is uncomfortable but straightforward for investors. In this cycle, owning the companies that literally build the AI future, the semiconductors, fiber, cooling systems, and specialized chips, has offered far better protection than betting on the software layer that may soon face disintermediation by the very technology it helped enable.

The broader implication is that 2026 may be remembered as the year the AI trade split into two very different stories: infrastructure as a near-term cash machine and software as a sector forced to prove its relevance in an age of intelligent automation. Wall Street is currently voting with capital, and that vote is heavily favoring the builders of the machine over the applications running on it.

A federal appeals court in Washington, D.C., on Wednesday denied Anthropic’s emergency request to temporarily halt the Department of Defense’s designation of the AI company as a supply chain risk, dealing a setback to the startup as its legal battle with the Trump administration continues.

The U.S. Court of Appeals for the D.C. Circuit ruled that the “equitable balance” favors the government. In pointed language, the panel stated: “On one side is a relatively contained risk of financial harm to a single private company. On the other side is judicial management of how, and through whom, the Department of War secures vital AI technology during an active military conflict.”

For that reason, the court denied the motion for a stay pending full review on the merits.

The decision creates a split outcome across courts. In a separate but related case last month, a federal judge in San Francisco granted Anthropic a preliminary injunction blocking the administration from enforcing a broader ban on the use of its Claude model across federal agencies. As a result, Anthropic remains excluded from direct DOD contracts and work involving defense contractors on Pentagon projects, but it can still serve other government agencies and non-defense customers.

Defense contractors are barred from using Claude in their DOD-related work but may continue using it for commercial or non-military projects.

The Pentagon labeled Anthropic a supply chain risk in early March 2026 under authorities including 10 U.S.C. § 3252 and provisions of the Federal Acquisition Supply Chain Security Act. The designation, historically applied mainly to foreign adversaries, requires contractors to certify they are not using Anthropic’s Claude models in military-related activities.

Defense Secretary Pete Hegseth first signaled the move publicly on X in late February, and President Trump followed with a Truth Social post directing agencies to immediately cease use of the technology, with a six-month phase-out period for certain systems.

Anthropic fired back with lawsuits filed in March, arguing the designation was retaliatory, unconstitutional, arbitrary and capricious, and procedurally flawed. The company contends the action punishes it for refusing to grant the Pentagon unrestricted access to Claude for “all lawful purposes.”

Anthropic had sought two narrow exceptions during contract talks: prohibiting use in fully autonomous lethal weapons and mass domestic surveillance of American citizens. Negotiations broke down after Anthropic held firm on these ethical red lines, despite having signed a $200 million contract with the Pentagon in July 2025 and previously deploying models on classified networks.

The appeals court acknowledged that Anthropic “will likely suffer some degree of irreparable harm” without a stay but characterized the harm as “primarily financial in nature.” It rejected the company’s First Amendment claims, noting that Anthropic failed to show its speech had been chilled during the litigation. Still, the court ordered a “substantial expedition” of the case given the stakes.

An Anthropic spokesperson said in a statement that the company is grateful the court recognized these issues need to be resolved quickly, and expressed confidence that “the courts will ultimately agree that these supply chain designations were unlawful.”

“While this case was necessary to protect Anthropic, our customers, and our partners, our focus remains on working productively with the government to ensure all Americans benefit from safe, reliable AI,” the statement said.

Acting U.S. Attorney General Todd Blanche hailed the ruling on X as “a resounding victory for military readiness.” He wrote: “Military authority and operational control belong to the Commander-in-Chief and Department of War, not a tech company.”

The dispute highlights deepening tensions between frontier AI labs and the national security establishment over control, safety guardrails, and the military applications of rapidly advancing technology. Anthropic, led by CEO Dario Amodei, has positioned itself as a leader in responsible AI development, emphasizing constitutional AI principles and caution around high-risk uses.

Critics inside the administration viewed the company’s stance as overly restrictive or even arrogant, especially amid ongoing military conflicts where AI tools could provide strategic advantages.

The case also raises broader questions about government leverage over private AI developers. The supply chain risk tool, designed to protect against vulnerabilities in critical supply chains, had rarely—if ever—been wielded against a major domestic technology firm before Anthropic.

Legal experts and nearly 150 former judges filed amicus briefs supporting Anthropic, arguing the designation bypassed required procedures and risked chilling innovation.

For now, the split rulings mean Anthropic faces real but contained restrictions: lost DOD revenue opportunities and potential reputational damage in defense circles, while retaining access to much of the federal government and the vast commercial market. The company has warned internally that prolonged blacklisting could cost billions in 2026 revenue.

The litigation is expected to move quickly. With the D.C. appeals court calling for expedition and the underlying merits still to be decided, the standoff could shape how other AI companies negotiate with the Pentagon.