Intel stock is swinging wildly, Dell just hit new highs, and AMAT reports earnings today. Here’s whether the AI hardware supercycle still has legs. Six months ago, “AI hardware trade” mostly meant Nvidia.

Problem: now the rally has spread — violently — into names that were left for dead just a year ago. Agitate: Intel stock is up over 300% in twelve months but just fell more than 30% from its June peak in a matter of weeks, which is either a warning sign or a buying opportunity depending on who you ask. Solution: breaking down Intel, Dell stock, and AMAT stock price action separately — rather than lumping them into one “AI trade” — reveals which parts of this rally are backed by real demand and which are running on sentiment. This matters right now because Applied Materials reports fiscal Q3 earnings today, August 13, a print the whole semiconductor equipment sector is watching.

Intel: Volatile Comeback or Overextended?

Intel has been the market’s most talked-about turnaround story, and the price action shows it:

• Shares traded near $101 this week, down from a 52-week high of $142.35 in June, but still up roughly 335%+ over the past year
• On August 10, Intel launched a $15 billion stock offering, diluting existing shareholders to fund its foundry ambitions
• CNBC’s Jim Cramer has publicly flagged Intel under CEO Lip-Bu Tan as a “focus name,” citing the foundry turnaround narrative

The read: Intel’s rally reflects real optimism about its foundry business and CHIPS-era manufacturing bets, but the recent 30%+ pullback shows how quickly sentiment can reverse when a name has run this hot.

Dell Stock: Quietly Making New Highs

While Intel grabs headlines, Dell stock has been the steadier AI infrastructure story:

• Shares closed near $505, up over 20% in a single session on record demand for AI-optimized servers
• Dell’s AI server order backlog hit a record $51.3 billion, with AI server revenue reaching $16.1 billion in its most recent quarter
• The stock has roughly tripled year-to-date

Why it’s different from Intel: Dell’s move is backed by an actual, quantifiable order backlog rather than a turnaround narrative — arguably a more durable signal.

AMAT: The Equipment Bellwether Reporting Today

AMAT stock price action has tracked the broader “picks and shovels” thesis of the AI buildout:

• Shares have gained roughly 195% year-over-year
• HSBC recently raised its price target to $683 from $522, maintaining a Buy rating
• Analysts expect Q3 revenue of about $8.99 billion, up roughly 23% year-over-year, in results due after today’s close

What to watch: Applied Materials sells the machines that make chips, not the chips themselves — its guidance is often read as a preview of demand across the entire semiconductor supply chain, including for Intel’s foundry ambitions.

Is the Hardware Supercycle Still Alive?

• Yes, structurally — order backlogs at Dell and capital spending commitments across the sector point to real, multi-year demand
• But not without volatility — Intel’s 30%+ round-trip in weeks shows how sentiment-driven parts of the rally remain
• AMAT’s earnings today will be a near-term litmus test for whether equipment demand is still accelerating or beginning to normalize

Actionable Takeaway

For your portfolio: treat Intel, Dell, and AMAT as three different bets, not one “AI hardware” basket. Dell’s backlog-driven strength and AMAT’s equipment-demand exposure represent more measurable fundamentals than Intel’s turnaround-and-dilution story. Watch today’s AMAT print closely — a soft guide could ripple across the entire chip-equipment complex within hours.

On its first day of trading on Shanghai’s STAR Market, ChangXin Memory Technologies — known globally as CXMT — did something few companies of any nationality have ever managed: it became more valuable than one of its country’s largest state-owned banks within hours of going public. Shares surged approximately 466% above their IPO price, closing the day with a market capitalization of roughly RMB 3.3 trillion — enough to overtake Industrial and Commercial Bank of China as the most valuable China-listed company, with more than RMB 140 billion of shares changing hands during the session alone.

The Numbers Behind the Debut

CXMT’s offering was not a marginal listing padded by speculative retail enthusiasm — it was Asia’s largest IPO of 2026 by a wide margin, and mainland China’s second-largest ever, trailing only Agricultural Bank of China’s $22.1 billion 2010 offering. The Hefei-based chipmaker raised 57.92 billion yuan, roughly $8.6 billion, pricing shares at 8.66 yuan before they closed the debut session at 49 yuan. Based on 2025 sales figures cited in its own IPO prospectus, CXMT held a 7.67% share of the global DRAM memory-chip market — positioning it as a genuine, if still distant, challenger to the three companies that have long dominated the sector: Samsung Electronics, SK Hynix, and Micron Technology.

The company’s underlying financials help explain investor enthusiasm. CXMT’s revenue reached 50.8 billion yuan, approximately $7.5 billion, in the first quarter of 2026 alone — a year-on-year increase of more than 700%, driven by surging AI-related demand for the DRAM chips used across AI servers, personal computers, smartphones, and automotive electronics.

Why This Listing Is a Geopolitical Story, Not Just a Financial One

Brookings Institution fellow Kyle Chan, an expert in China’s technology policy, framed CXMT’s significance in explicitly strategic terms, describing the company as playing a critical role in China’s AI push, particularly in the face of US export controls. That framing matters because of what US restrictions specifically target: Washington’s export-control regime has barred China from importing high-bandwidth memory (HBM) chips — a high-performance category of DRAM that is critical for training and running advanced AI models. CXMT’s expansion is, in effect, China’s most concrete industrial answer to that restriction: building domestic capacity in the exact chip category the US has tried hardest to keep out of Chinese hands.

The timing also matters. CXMT’s Shanghai debut followed closely behind South Korea’s SK Hynix completing a $26.5 billion Nasdaq IPO, meaning global capital markets absorbed two of the memory-chip industry’s largest-ever public offerings within weeks of each other — a signal of just how central memory chips have become to the broader AI infrastructure investment cycle reshaping capital markets globally in 2026.

The Cash-Drain Concern

Not every signal ahead of the listing was unambiguously bullish. In the days before the debut, CXMT’s looming IPO stoked fears of a broader cash drain from Chinese equities, as investors pulled capital from other Chinese tech holdings to fund participation in what was widely expected to be an oversubscribed offering — a dynamic that contributed to a pullback in Chinese technology shares in the sessions immediately preceding the listing.

Analysts have also flagged sustainability questions about the memory sector’s current earnings profile more broadly. One market strategist cautioned that the industry may be nearing a short-term peak in memory-cycle sentiment, warning that the exceptional margins and profitability currently visible across the DRAM sector are unlikely to persist through a full cycle and will eventually normalize — a caution that applies to CXMT’s own trajectory as much as to its global peers, even as the company’s near-term revenue growth remains extraordinary.

What Comes Next for China’s Chip Ambitions

CXMT has stated its IPO proceeds will be deployed primarily toward mass production of memory wafers and expanded R&D — a direct, capital-intensive bet on scaling output rather than diversifying into adjacent businesses. For China’s broader semiconductor self-sufficiency strategy, CXMT’s success (or eventual stumble) will serve as a bellwether for whether domestic Chinese chipmakers can translate state-backed capital access and captive domestic demand into genuine competitiveness against entrenched South Korean and American incumbents — the same question underlying Beijing’s parallel investments across the semiconductor supply chain, from lithography equipment to rare-earth-dependent chip materials.

The Bottom Line

CXMT’s 466% debut is simultaneously a financial-markets story, an AI-infrastructure story, and a geopolitics story — and the three are now inseparable. For investors and policymakers tracking the broader US-China technology competition, CXMT’s post-IPO performance over the coming quarters will offer one of the clearest available signals of how effectively Chinese state-directed capital can compensate for continued exclusion from the most advanced Western and allied semiconductor technology.

Inside the sprawling halls of Paris’s Porte de Versailles, the atmosphere at the tenth anniversary of Europe’s premier technology gathering feels remarkably sober. The flashing holograms and robotic dogs of previous years have been quietly pushed to the periphery. Instead, the defining VivaTech AI trends centre on something far less cinematic: immediate, measurable commercial utility. Ten years since its inception, the conference has outgrown its adolescent fascination with what technology could do. Now, European founders and international investors are betting everything on artificial intelligence that actually works on the factory floor, in the back office, and across the supply chain.

This shift at VivaTech mirrors a broader correction across the global technology sector. The initial speculative frenzy surrounding generative models has collided with the harsh realities of corporate budgets and data privacy constraints. We have officially entered the deployment phase. Executives no longer want to pay for experimental software that hallucinates legal precedents or hallucinates customer service responses. They demand secure, ring-fenced tools that drive margin expansion.

The numbers reflect this systemic maturation. According to recent data synthesized by the Organisation for Economic Co-operation and Development (OECD), enterprise adoption of applied AI models is projected to drive a 1.4% annual increase in labour productivity across the Eurozone by 2027. Yet, the same dataset reveals a glaring friction point: only 18% of mid-sized firms have successfully integrated these models beyond pilot programs.

The gap between pilot and production is where the money is now being made. European venture capital has adjusted its focus accordingly. According to the Financial Times, funding for pure-play foundation model startups dropped by 22% in the first quarter of 2026, while capital allocated to vertical-specific AI applications surged. Investors are no longer funding the picks and shovels; they are funding the extraction.

Walking the convention floor this May, the changing guard is impossible to ignore. Startup booths are stripping the phrase Large Language Models (LLMs) from their primary marketing copy. The pitches have transformed. Founders are no longer selling the intelligence of their neural networks; they are selling automated invoice reconciliation, predictive supply chain routing, and immediate cost reduction.

Arthur Mensch, CEO of Paris-based Mistral AI, summarised this shift during a closed-door briefing on Tuesday. He noted that enterprise clients have abandoned open-ended experimentation in favour of strict, highly defined use cases. This pragmatism is fundamentally reshaping the European tech ecosystem. The continent, long criticised for failing to produce consumer internet giants, is leaning heavily into its traditional strengths: industrial engineering, regulatory compliance, and complex B2B software.

The capital backing these ventures is equally pragmatic. The French state investment bank, Bpifrance, announced a €500 million facility specifically earmarked for enterprise AI adoption within legacy manufacturing firms. This is not speculative capital. It is modernisation infrastructure. By targeting established industries, European policymakers are attempting to engineer an economic transition rather than merely chasing Silicon Valley’s consumer-focused tail.

That said, selling applied intelligence requires an entirely different sales motion. Startups must now prove integration capabilities with legacy SAP and Oracle databases. They have to navigate complex procurement cycles. The romantic era of the overnight AI unicorn is dead. We are now in the era of the gruelling enterprise sales cycle, where security audits matter more than parameter counts.

This transition toward utility is not happening in a vacuum. It is being heavily engineered by Brussels. The enforcement of the European AI Act has fundamentally altered the structural economics of software development on the continent. Critics initially warned that the legislation would stifle innovation, but the reality on the ground at VivaTech suggests a different outcome. Regulation has inadvertently created a massive market for compliance-grade, sovereign AI solutions.

What are the main AI trends at VivaTech?

At VivaTech, the primary AI trends centre on applied artificial intelligence, strict regulatory compliance under the EU AI Act, and enterprise-grade deployment. Companies are actively abandoning generative novelty in favour of measurable productivity gains, secure sovereign data solutions, and demonstrable return on investment.

This compliance-first approach offers a distinct competitive moat. American tech giants are currently battling copyright infringement lawsuits and regulatory scrutiny regarding their data scraping methodologies. European startups, conversely, are building models explicitly trained on licensed, opt-in data. They are offering guarantees that foreign competitors cannot match. When a German automotive manufacturer integrates a predictive maintenance model, they require absolute certainty that their proprietary telematics data will not be used to train a public model.

The picture is more complicated than a simple trans-Atlantic rivalry. It is a divergence in product philosophy. The US model prioritises general intelligence and rapid consumer adoption. The emerging European model, showcased vividly across the VivaTech pavilions, prioritises domain-specific accuracy, data sovereignty, and legal safety. In the enterprise sector, safety is rapidly becoming a premium feature rather than a bureaucratic burden.

The downstream consequences of this shift are profound for both policymakers and small-to-medium enterprises (SMEs). For the latter, the barriers to entry are finally lowering. For the last three years, AI deployment was effectively restricted to multinational corporations with vast engineering resources. The current generation of applied tools, heavily promoted at VivaTech, operates as plug-and-play software.

This democratization of capability will aggressively disrupt traditional B2B service sectors. Legal research, entry-level accounting, and supply chain logistics are facing immediate margin compression. According to a recent analysis by Bloomberg Intelligence, professional services firms that fail to adopt automated workflows will see their operating margins contract by up to 15% over the next 24 months. The cost of remaining analogue is becoming fatal.

Still, this transition requires massive infrastructure. The bottleneck has shifted from software capability to physical compute. Sovereign data solutions demand localized data centres. European nations are currently scrambling to build the requisite energy and cooling infrastructure to support this localized compute demand. The next major geopolitical battleground will not be the algorithms themselves, but the raw gigawatts required to run them domestically.

Governments are acutely aware of this vulnerability. French President Emmanuel Macron used his opening address at the conference to announce accelerated permitting processes for green-energy data centres. The goal is clear: to ensure that the intellectual property generated by European applied AI remains physically housed within the borders of the European Union.

Competing Perspectives: The Compute Deficit and Market Fragmentation

Not everyone in the halls of Porte de Versailles shares this optimistic vision of a European industrial renaissance. A vocal contingent of investors argues that the continent’s focus on applied AI is essentially a concession of defeat in the foundational model race. The bear case is structural and compelling.

Europe remains fragmented. A startup cannot scale across the continent without navigating 27 different legal jurisdictions and language barriers. More critically, the hardware deficit is severe. According to Reuters technology analysts, Europe currently accounts for less than 12% of the global advanced GPU supply. You cannot build a sovereign AI ecosystem if you rely entirely on Californian hardware manufactured in Taiwan.

Dissenting voices argue that by focusing purely on B2B applications, European firms risk becoming entirely dependent on the foundational API layers controlled by OpenAI, Google, and Anthropic. If the base cost of inference rises, the profit margins of these European applied AI companies will collapse. In this view, the regulatory moat created by the AI Act is a temporary illusion, easily breached once the foundational models reach a threshold of undeniable superiority.

Yet, the counter-argument remains potent. Foundational models are rapidly commoditising. Open-source alternatives are narrowing the performance gap weekly. If intelligence becomes a cheap, ubiquitous utility, the real economic value will accrue to the companies that own the proprietary workflow integrations and the industry-specific data.

Ten years on, VivaTech has shed its adolescent idealism. The focus on artificial intelligence that practically functions within the rigid constraints of modern business represents a necessary maturation. Europe is no longer attempting to clone Silicon Valley. It is building an ecosystem tailored to its own industrial and regulatory DNA.

The tension between foundational dependence and applied utility will define the next decade of enterprise technology. However, the mood in Paris suggests a quiet confidence that the pendulum is swinging back toward business fundamentals. The era of the speculative demo has officially concluded; the era of ruthless execution has begun.