On July 14, 2026 (Beijing time), IBM made an unusual move by releasing preliminary second-quarter results a full week ahead of its scheduled earnings announcement. The following day (July 15), IBM’s stock plunged 25.21%, closing at $217.07, with trading volume surging to $14.8 billion. This marked IBM’s largest single-day drop since trading records began in 1968, surpassing even the 23.7% decline during the "Black Monday" crash of 1987.
But IBM’s historic plunge wasn’t an isolated event for the century-old tech giant. On the same trading day, NVIDIA ended up 4.06% at $211.80, Micron soared 4.92% to $983.12, and SK Hynix ADR skyrocketed 27.3% to $193.92, setting a new all-time high. Hardware and software, chips and services—stock prices traced sharply divergent paths along the same timeline.
This wasn’t a random swing in market sentiment, but a clear structural signal: capital spending on AI is shifting rapidly from software and services to hardware infrastructure. Starting with the financial details behind IBM’s plunge, this article will break down the logic driving this shift, analyze how hardware supply chain giants like NVIDIA, TSMC, Broadcom, Micron, and SK Hynix are benefiting, and finally examine whether IBM can avoid becoming just another "legacy software company" in the AI era.
The Real Reason Behind IBM’s 25% Plunge: Corporate AI Budgets Are Being Reallocated
Financial Data: Revenue and Profits Both Disappoint
According to IBM’s preliminary Q2 2026 financials, the company posted $17.2 billion in revenue, up just 1% year-over-year. Analysts had expected $17.86 billion, so actual revenue missed forecasts by about 4%. Adjusted EPS came in at $2.93, also below the market expectation of $3.02.
Breaking it down by business segment: software revenue grew 5% year-over-year, well below the expected 11% growth; consulting was essentially flat, up only 1% at constant currency; infrastructure revenue fell 7% year-over-year, worse than the expected 3% decline. The infrastructure division, which includes mainframe sales, was hit hardest, with sales down 7%. UBS and Goldman Sachs analysts estimate that IBM’s mainframe (zSeries) Transaction Processing revenue dropped by double digits year-over-year, and this segment accounts for nearly 30% of IBM’s software revenue.
CEO’s Statement: Four Major Challenges and One "Unanticipated" Shift
IBM Chairman and CEO Arvind Krishna openly acknowledged four major challenges in his letter to investors:
First, z17 mainframe sales fell short of expectations, dragging down related software sales, especially in transaction processing.
Second, in the last weeks of June, clients abruptly shifted their capital spending toward servers, storage, and memory, anticipating imminent price hikes and wanting to lock in supply before prices rose. Krishna admitted, "We anticipated some supply chain impact, but failed to foresee the extent of capital spending reprioritization."
Third, rapidly escalating cybersecurity concerns across the industry distracted clients from procurement.
Fourth, execution issues—Krishna conceded, "These conditions required flawless execution from our team, but we fell short." The team’s adaptation and response were too slow, causing many large deals to miss their deadlines.
Of these four challenges, the second is the most critical. It doesn’t point to IBM’s own strategic missteps or product flaws, but rather signals a structural shift across the entire industry: enterprise clients are reallocating IT budgets away from software and services toward hardware infrastructure on a large scale.
From "Software First" to "Hardware Rush": The Logic Behind Budget Flows
IBM’s earnings warning wasn’t unique. On the same trading day, software stocks broadly weakened: ServiceNow fell 5.8%, Workday dropped 3.5%, SAP declined 3.2%, and Salesforce slid 2.1%. Meanwhile, AI hardware and chip stocks surged.
This divergence reflects rational choices by CFOs and CIOs working with limited budgets. As Morningstar analyst Luke Yang noted, IBM’s earnings "aren’t just a single company’s stumble—they highlight how the AI investment boom is reshaping enterprise IT spending. With limited tech budgets, massive funds are flowing to hardware companies, leaving little for other tech sectors." CNBC’s "Mad Money" host Jim Cramer summed up the trend as three top priorities for enterprise IT budgets—cybersecurity, AI hardware, and AI compute costs—with spending outside these categories increasingly delayed or sidelined.
Goldman Sachs’ immediate research report bluntly stated that IBM’s event "fully validates the software bear case," predicting "broad" downward pressure on software and services. Barclays analyst Andrew Keches pointed out that IBM management had repeatedly emphasized AI’s "additive" effect on the software stack, not a destructive one, but this earnings gap centered on mainframes and related Transaction Processing software, with clients shifting spending to scarce servers, storage, and memory—making the previous narrative untenable.
Zooming out, Goldman Sachs projects that global AI capital expenditures on compute, data centers, and power will reach about $7.6 trillion from 2026 to 2031, with annual investment rising from $765 billion in 2026 to $1.64 trillion in 2031. Citi has raised its forecast for total AI industry revenue from $2.8 trillion to $3.3 trillion for 2026–2030, and capital spending from $8.0 trillion to $8.9 trillion in the same period. This enormous flow of capital is now pouring into hardware infrastructure at an unprecedented pace, not software subscriptions.
Why Are NVIDIA, TSMC, Broadcom, Micron, and SK Hynix Still Benefiting?
When enterprise AI budgets shift from software subscriptions to hardware procurement, the capital flow is clear and traceable. Here’s how five core hardware suppliers are benefiting, along with their latest financial performance.
NVIDIA: The Core Supplier for GPU Demand
NVIDIA is the most direct beneficiary of this AI capital spending shift. On July 15, NVIDIA closed at $211.80, up 4.06% in a single day. In Q2 2026, NVIDIA’s net income hit $26.42 billion, up 59.2% year-over-year; operating cash flow reached $15.37 billion, up 6.1%. Analysts’ average price target for NVIDIA is $309.78.
NVIDIA’s value proposition is straightforward: regardless of which cloud provider enterprises choose or what AI application stack they build, they ultimately need GPUs for model training and inference. NVIDIA’s H100 and B200 GPUs have become the "standard configuration" for AI infrastructure.
TSMC: The Key Node for AI Chip Manufacturing
TSMC (TSM) ADR closed at $420.39 on July 15, with trailing twelve-month EPS of $11.51 and a P/E ratio of about 36.52. As the world’s most advanced semiconductor foundry, TSMC handles the vast majority of advanced process orders for major AI chipmakers like NVIDIA, AMD, and Broadcom. Whether it’s GPUs, AI networking chips, or custom ASICs, manufacturing ultimately relies on TSMC’s cutting-edge process nodes. Every expansion in AI compute demand directly translates into increased wafer shipments for TSMC.
Broadcom: The Key Supplier of AI Networking Chips
Broadcom (AVGO) closed at $389.11 on July 15, up 1.32%. In Q2 2026, Broadcom’s revenue grew 47.9% year-over-year, operating income reached $10.8 billion (up 85%), and free cash flow hit $10.3 billion (up 60.1%).
Broadcom’s advantage lies in AI data center networking infrastructure. As GPU clusters scale from thousands to tens of thousands—even hundreds of thousands—of cards, inter-GPU networking becomes the critical bottleneck for cluster efficiency. Broadcom holds a central position in AI networking chips (such as Ethernet switch chips and PCIe switch chips), making it indispensable for AI data center networking.
Micron: Explosive Demand for HBM Memory
Micron (MU) surged 4.92% on July 15, closing at $983.12 with trading volume reaching $29.718 billion—the highest on the day among US stocks. In Q2 2026, Micron’s revenue reached $23.86 billion, up 196.3% year-over-year; gross profit hit $17.76 billion, up 499%.
Wall Street forecasts that Micron and the broader chip industry will generate about $700 billion in profits in 2027. Micron’s net income was just $9 billion in fiscal 2025, but is expected to soar to $83 billion in 2026 and further to $176 billion in 2027.
Micron’s explosive growth is driven directly by demand for HBM (High Bandwidth Memory). AI training requires massive high-bandwidth memory alongside GPUs, and HBM has become the second most expensive component in AI servers after GPUs. As one of the world’s top three HBM suppliers, Micron is fully benefiting from this structural demand surge.
SK Hynix: The Leading Player in the HBM Market
SK Hynix ADR soared 27.3% on July 15, closing at $193.92—a new all-time high. Intraday, it rose over 22% to $186.24, with total market cap reaching $1.36 trillion.
The immediate catalyst for this surge was a bullish report from top research firm SemiAnalysis titled "Be Greedy When Others Are Fearful," which explicitly favored SK Hynix. SK Hynix leads Micron in HBM market share and is a major HBM supplier for NVIDIA GPUs. As AI training demands ever greater memory bandwidth, the pace of HBM iteration and capacity expansion directly determines SK Hynix’s growth potential.
Will IBM Become a Laggard in the AI Era?
Existing Strengths: Red Hat, AI Services, and Long-Term Quantum Computing Investments
IBM isn’t without assets. In its software business, Red Hat’s revenue growth accelerated to 11%, and recent acquisitions HashiCorp and Confluent are performing strongly. Distributed infrastructure grew 37%, with backlog orders around $500 million. On July 8, IBM announced Lightwell—a $5 billion AI capability commitment, staffed by over 20,000 engineers, with early adopters including Bank of America, Citi, Goldman Sachs, JPMorgan Chase, Mastercard, Morgan Stanley, and Wells Fargo.
In quantum computing, IBM announced a partnership with the US Department of Commerce to build the Anderon quantum wafer foundry, securing $1 billion in CHIPS Act incentives. IBM is committing another $1 billion in cash and plans to invest over $10 billion in quantum over the next five years. IBM has maintained dividend payments for 56 consecutive years, with a current yield of 2.33%.
Structural Challenges: On the Wrong Side of the Capital Flow
However, these long-term initiatives are unlikely to offset the current structural pressures in the short term. IBM’s predicament is that its core revenue sources—mainframe software licensing, transaction processing systems, and enterprise consulting—are at the bottom of enterprise budget priorities.
As Cramer pointed out, IBM "has too many products and services falling into the ‘other spending’ bucket—even if they have a decent AI story to tell." When a client CFO must choose between "buying IBM software" and "rushing to secure AI servers and storage," the urgency of the latter overwhelmed everything in the closing weeks of June.
Barclays analyst Andrew Keches offered a sharper observation: the market’s central debate is whether this is a "temporary misalignment" in enterprise purchasing, or the early sign that AI infrastructure investment is "systematically crowding out" traditional software spending. If it’s the latter, IBM faces not just a one-off quarterly miss, but a multi-year restructuring of its revenue mix.
Investment Perspective: What IBM Needs to Prove
From an investment standpoint, IBM’s current market cap is about $272.8 billion, with a P/E ratio of 25.7 and a PEG ratio of 0.27. A low PEG signals that the market doesn’t expect much growth—a warning in itself.
IBM’s core challenge is to prove to the market that it can transform from a "legacy software and services company" into an AI-era infrastructure player. This requires breakthroughs on two fronts: first, shifting more revenue weight to open-source and cloud-native assets like Red Hat and HashiCorp, reducing reliance on mainframe-related income; second, scaling AI services like Lightwell from "early adoption" to "widespread deployment," making them a real part of enterprise AI budgets.
If IBM can’t show clear revenue mix changes in the next two to three quarters, the market’s pricing logic may shift from "a tech giant in transition" to "a legacy software company left behind by the AI era."
Conclusion
IBM’s record-setting 25% plunge on July 15, 2026, wasn’t just a company’s misstep—it was a concentrated signal of structural shifts in AI capital spending. Funds are flowing away from software subscriptions, enterprise consulting, and mainframe transaction processing, toward GPUs, wafer manufacturing, AI networking chips, and high-bandwidth memory.
The stock price and earnings growth of NVIDIA, TSMC, Broadcom, Micron, and SK Hynix clearly trace the path of this capital flow. Meanwhile, the collective pressure on IBM, Salesforce, Workday, and other software and services companies outlines the other side of this trajectory.
For investors, understanding the duration and depth of this shift matters far more than whether IBM beats quarterly expectations. If this is just a temporary procurement timing issue, software stock valuations may soon recover; but if it marks the beginning of AI infrastructure spending systematically crowding out traditional software, the investment paradigm for tech stocks faces a profound repricing.
IBM’s next steps—whether Red Hat and Lightwell can truly become growth engines—will largely determine whether this century-old company can avoid becoming a "legacy software company" in the AI era.
FAQ
Q: What was the direct cause of IBM’s 25% stock plunge?
IBM released preliminary Q2 results a week before its official July 22 earnings announcement: revenue of $17.2 billion missed the $17.86 billion forecast, and adjusted EPS of $2.93 missed the $3.02 forecast. CEO Arvind Krishna admitted that in the final weeks of June, clients shifted capital spending en masse to servers, storage, and memory, causing several large software and mainframe deals to fall through.
Q: Why are hardware companies like NVIDIA and TSMC benefiting instead?
Enterprise AI budgets are moving from software subscriptions to hardware infrastructure. NVIDIA provides the GPUs essential for AI training and inference; TSMC is the manufacturing core for AI chips; Broadcom supplies AI data center networking chips; Micron and SK Hynix are leading suppliers of HBM high-bandwidth memory. Every AI hardware purchase flows directly into these companies’ revenue.
Q: Will software stocks remain under pressure?
Goldman Sachs has warned that the IBM event "fully validates the software bear case." The key depends on whether the budget shift to hardware is just a short-term adjustment or a long-term structural change. If it’s the latter, software and services will face sustained capital outflows.
Q: Does IBM still have a chance to recover?
IBM still holds quality assets like Red Hat (11% growth), HashiCorp, Lightwell ($5 billion AI services commitment), and long-term quantum computing investments. But the key is whether these new businesses can take on a larger share of revenue within two to three quarters, proving IBM isn’t just a "legacy software company left behind by the AI era."
Q: How big is AI capital spending?
Goldman Sachs estimates global AI capital spending will total about $7.6 trillion from 2026 to 2031, with annual investment rising from $765 billion in 2026 to $1.64 trillion in 2031. Citi expects total AI industry revenue of about $3.3 trillion and capital spending of about $8.9 trillion over the same period.
