Panasonic’s Three-Year AI Infrastructure Wager Hits ¥500 Billion

Panasonic Holdings, the Japanese electronics conglomerate once defined by consumer appliances, is staking its next chapter on AI infrastructure. A year-long surge in demand for the company’s batteries, circuit-board materials and capacitors used inside AI data centers has more than doubled its stock, lifting it to a record valuation of ¥11.5 trillion (US$71 billion), with shares at their highest since September 1974.

Chief executive Yuki Kusumi told reporters in a July 1 roundtable that the company would pour around ¥500 billion into AI-infrastructure businesses over the next two fiscal years, targeting approximately ¥1.4 trillion in AI-related sales by fiscal 2029. The wager is the clearest expression yet of a pivot Kusumi has been building since he took the top job at the holding company, and it lands as Panasonic’s appliances business has shrunk from the dominant force it once was. The targets are tied to specific Awards from cloud customers, not to a broad consumer-market recovery. Storage batteries, multilayer PCB materials, and capacitors sit on the critical path of every hyperscaler building AI data centers today.

The Bet on AI Infrastructure

Panasonic’s ¥11.5 trillion valuation is itself the headline. It is the highest market value for the company in any month since market data begins in September 1974, the earliest date for which figures are available. That re-rating reflects a market view that Panasonic’s components, long treated as a slow-moving industrial supplier, now sit on the critical path of the AI buildout.

For Kusumi, the three-year horizon is the heart of the strategy. In the July 1 roundtable, he laid out a six-year arc: the first half to capture the AI-infrastructure opportunity, the second to make it the next engine of growth. The framing matches a broader Panasonic restructuring of its appliance R&D labs, where AI-guided robots now feed disassembly data back into product design, and the AI repositioning has touched nearly every layer of the company’s hardware. The wager is whether an appliance-era conglomerate can be rebuilt as an AI-era components supplier while its customers’ data-center demand is still climbing.

What Three Years and ¥1.4 Trillion Buys

In May 2026, Panasonic published its Group Growth Strategy to 2032, replacing the conventional three-year mid-term plan it had used for decades. The new document sets targets for the AI-infrastructure businesses in concrete figures, according to the Group Growth Strategy interview from May 2026. The company expects AI-infrastructure sales to reach approximately 1.4 trillion yen in fiscal 2029, alongside 290 billion yen in adjusted operating profit from the same businesses. The cumulative investment of around 500 billion yen between fiscal 2027 and fiscal 2029 will fund capacity expansions and next-generation products.

The full Group path runs through three phases. Phase 1 stretches from the current fiscal year through fiscal 2029, with the Devices area, the unit that makes PCB materials, capacitors and battery cells, doing the heavy lifting. Adjusted operating profit for the whole Group is targeted at 600 billion yen in fiscal 2027 and more than 750 billion yen in fiscal 2029.

The targets sit alongside a separate, older ambition Panasonic set in January 2025: lift AI-driven hardware, software and solution businesses to 30 percent of revenue by 2035.

The Hardware Stack Behind AI Servers

The Devices-area work is not one product but a stack. Inside a generative-AI data center, Panasonic supplies components for the GPU and ASIC circuits, the power systems that feed them, and the backup units that smooth out peaks when a thousand accelerators spin up at once. The PCB materials race inside AI servers is part of a global PCB supply squeeze reshaping AI hardware costs, which has lifted the price of high-end multilayer substrates in just a few months.

The mix matters because peak power has become the binding constraint on data-center builds. GPUs and ASICs in current AI servers draw power in sudden surges, and managing those surges lets operators avoid building excess grid capacity. Panasonic says its “Award” win rate, the share of customer development agreements it has secured for its data-center energy storage systems, has already reached 80 percent of its sales target for the AI-infrastructure business. Awards have been granted on nearly every product line included in the ¥1.4 trillion number. Panasonic’s pitch to hyperscalers sits on two engineering facts. The first is that AI accelerators draw power in sudden, large spikes that strain grid capacity; the second is that lithium-ion storage systems and conductive polymer capacitors can absorb those spikes faster than grid-scale infrastructure can be expanded.

• Data-center energy storage systems: Panasonic Energy’s lithium-ion based systems used for backup power and peak suppression at hyperscaler sites
• High-performance multilayer PCB materials: Panasonic Industry substrates designed to handle the signal speeds and heat loads of GPU, ASIC and CPU circuits
• Conductive polymer capacitors: components that maintain capacitance in the high-temperature environments around accelerators
• Battery backup units (BBUs): clusters of cells that absorb sudden load swings at AI server racks
• Capacitor backup units (CBUs): fast-discharge units that handle millisecond-scale power events that BBUs are too slow to catch

Restructuring to Fund the Wager

Funding the new push requires cutting somewhere else. In May 2026, Panasonic expanded the scale of planned job cuts to as many as 12,000 positions, up from the 10,000 it had previously signaled. The expansion lifts expected annual savings to ¥145 billion compared with fiscal 2024, money that the holding company says it will redirect into AI-infrastructure capex. The job cuts span fiscal 2025 and fiscal 2026, according to the company’s restructuring plan.

The restructuring reaches into the company’s software backbone. As part of the January 2025 Panasonic Go launch statement, the company said it had invested over US$10 billion in North America to acquire Blue Yonder, a supply-chain AI platform, and to build out its internal AI platform.

Internally, Panasonic has rolled out an AI assistant called PX-AI, powered by large language models, to approximately 180,000 employees. Job cuts and operating-software build-out are the two prongs that fund and operationalize the AI infrastructure push. The shift from hardware-only legacy businesses to AI-driven services is the second half of the same pivot Kusumi outlined in the Group Growth Strategy. The US$10 billion Blue Yonder investment is the company’s largest AI-platform bet to date. PX-AI is the most visible internal deployment of generative AI across the company’s white-collar workforce.

Redirecting a Tesla Battery Plant to AI Power Systems

The clearest physical expression of the wager is inside a US battery plant. Panasonic is seeking to redirect production at a US facility that had struggled as EV battery production for Tesla was cut back, converting the lines into a manufacturing base for AI power systems. The shift is a small number in dollar terms inside the ¥500 billion plan, but a symbolically large one for Panasonic’s relationship with Tesla, the customer that made its battery unit a global supplier.

The economics of AI power systems differ from EV cells in two ways. EV batteries are designed around energy density and long cycle life, while data-center backup units are optimized for repeated high-power discharge events, the spikes that hyperscalers need to absorb. Panasonic’s Group Growth Strategy explicitly describes energy storage systems for data centers as part of the “heart” of AI servers. Panasonic says its storage battery systems play a critical role in curbing peak power demand at AI data centers, a function that becomes more valuable as accelerator power consumption climbs with each chip generation.

The next three years will be measured against whether Panasonic can keep the win rate on Awards high enough to fill the ¥1.4 trillion revenue line, and whether the cohort of Japanese peers keeps lifting the countrywide case that AI demand is durable. The bet depends on a market, AI data-center power, that is itself shifting as hyperscalers move toward their own silicon, in-house power systems, and longer-duration energy storage chemistries. The Tesla-plant redirect, on its own, is a small enough line item that its failure would not derail the ¥500 billion plan. The strategy’s larger test sits in whether the Devices area can keep winning Awards at the current pace. Kusumi has flagged risk anticipation as a strategic priority, telling reporters the lead may not continue into the future.

For the next three years, we intend to capture this opportunity. The following three years will be about making that our next engine of growth.

Yuki Kusumi, Group CEO of Panasonic Holdings, in a July 1, 2026 roundtable interview.

The Anthropic Partnership and the 30% Target

The hardware push is paired with a software bet. In January 2025, Panasonic announced Panasonic Go, a corporate-wide initiative built around a global strategic partnership with Anthropic PBC, one of the US AI research companies leading the industry. The tie-up is the first major software-side commitment the appliance-era conglomerate has made in generative AI.

The deal sets an explicit revenue target: Panasonic said it would expand its AI-driven hardware, software and solution businesses to 30% of revenue by 2035, a ten-year arc that runs well beyond the three-year fiscal 2029 targets for AI-infrastructure hardware. The Anthropic deal is the consumer-facing side of the partnership, and Panasonic Well, the group’s Palo Alto-based wellness unit, was the first to embed Claude, Anthropic’s assistant, into a consumer product called Umi. The internal side uses Anthropic’s models inside PX-AI, the assistant now in front of roughly 180,000 employees, and as one of the layers in Blue Yonder’s supply-chain AI orchestration. Whether Anthropic’s models stay at the leading edge of the industry for the next decade is a separate risk that affects the 30% number more than the ¥1.4 trillion figure.

Both targets sit inside the same wager but measure different parts of the company: the ¥1.4 trillion goal tracks how fast Panasonic can build AI-infrastructure devices, and the 30% by 2035 target tracks how fast the same businesses can re-mix their revenue toward AI-related goods and services. The hardware side depends on Panasonic’s own factories; the software side depends on partners whose roadmaps the company does not control.

Why a 25-Times P/E Isn’t Reading as Expensive

The market’s verdict on Panasonic’s wager is now visible in its valuation. At least one sell-side analyst does not think the multiple is stretched. Norikazu Shimizu, an analyst at IwaiCosmo Securities, made the point that Panasonic’s multiple is not high compared with stocks whose earnings are more sensitive to AI infrastructure demand. The roughly 25-times P/E looks rich on its face, but Shimizu’s framing points to a structural reason: not all of Panasonic’s earnings power is AI-linked.

Panasonic still operates in home appliances, housing-related businesses, automotive operations and software. Shimizu’s argument is that the AI-exposed portion of the company’s earnings, if isolated, would carry an even higher implied multiple than the headline 25 times; the conglomerate structure is what holds the consolidated P/E at the 25-times level.

If Panasonic’s Devices area delivers on the ¥1.4 trillion target, the implied multiple on the AI portion of the business is much higher than the headline 25 times suggests. That math is what investors are paying for today, and it is also what Kusumi’s three-year window is meant to deliver. The P/E framing matters because the share price has already priced in a meaningful AI contribution; missing the ¥1.4 trillion number would force a multiple reset. For now, the market is willing to give Panasonic credit while it executes. The 80 percent Award win rate Panasonic cited for its data-center energy storage systems is what gives the market the conviction to pay for the AI contribution.

Panasonic’s current price-earnings ratio of around 25 times isn’t necessarily high compared with stocks whose earnings are more sensitive to demand for AI infrastructure.

Norikazu Shimizu, analyst at IwaiCosmo Securities, in coverage of the record-valuation move.

The Japanese Cohort Riding the Same Wave

Panasonic is not the only Japanese company benefiting from the AI buildout. Kioxia Holdings Corp, the NAND flash supplier, has seen its valuation climb as memory pricing has firmed on AI storage demand. Fiber-optic cablemaker Fujikura Ltd has also benefitted from robust sector demand, with its shares up sharply as hyperscalers build out the inter-rack and inter-data-center cabling that AI clusters require. The two companies anchor a broader cohort that includes the semiconductor production equipment leaders and the rare-earth materials suppliers that feed the data-center supply chain.

Recently, the Bank of Japan said that strong exports to the AI sector are providing an economic lift, the same observation that frames the broader Japanese equity story of 2026. The macro tailwind matters for Panasonic because parts of its growth story depend on Japan maintaining its position as a diversified supplier of AI components, not just a single-product vendor. Japan’s broader push to capture the AI buildout has lifted equipment makers, materials suppliers, and storage providers in a coordinated cycle.

The cohort effect is double-edged: when Japanese AI suppliers win, the country’s supply chain becomes a more attractive place for hyperscalers to source from, which lifts Panasonic’s pitch; when one stumbles on a product cycle, it can also drag the others into the same conversation about whether Japan’s AI exposure is structural or cyclical.

• Panasonic’s record valuation: ¥11.5 trillion (US$71 billion)
• Panasonic AI-infrastructure sales target: ¥1.4 trillion by FY2029
• Panasonic Group AOP target: more than 750 billion yen by FY2029
• Japanese suppliers named in the rally: Kioxia (NAND flash) and Fujikura (fiber-optic cable)
• Bank of Japan position: AI-related exports are providing an economic lift

Kusumi on What Could Break the Lead

Kusumi himself warned, in the same July 1 interview, that Panasonic’s current competitive advantage in storage batteries for AI data centers may not necessarily continue into the future. The remark was framed as risk anticipation, not as a forecast of imminent failure, but it tells readers that the company itself is treating the lead as conditional rather than durable. His language was direct: “We must assume that change is constant and be prepared to change in every direction, at any time.” That posture matters because the company has now committed ¥500 billion in capex and roughly 12,000 job cuts to a three-year bet that depends on a market, AI data-center power, that is itself shifting. Hyperscalers are designing their own silicon, building their own power systems, and exploring longer-duration energy storage chemistries.

Each of those moves chips away at the components market Panasonic is targeting. The company says its storage battery systems play a critical role in curbing peak power demand at AI data centers, but a customer that designs its own batteries and capacitors inside the same building does not need to buy them from Panasonic. The strategy also has to clear the cohort test: Kioxia and Fujikura have already shown that Japanese AI suppliers can scale quickly, and Panasonic’s wager is whether its own version of that story, told through storage batteries and PCB materials rather than memory or cabling, can hold through the next AI hardware cycle. Bank of Japan research has linked Japan’s AI exports to broader economic momentum, which lifts the cohort but does not insulate any single name from product-cycle risk.

Frequently Asked Questions

What is Panasonic’s AI infrastructure target?

Panasonic is targeting approximately 1.4 trillion yen in sales from AI-infrastructure businesses in fiscal 2029, with 290 billion yen in adjusted operating profit from the same segment. The figure covers data-center energy storage systems, high-performance PCB materials, capacitors and backup units, and was published as part of the Group Growth Strategy to 2032 in May 2026.

How is Panasonic funding its AI infrastructure bet?

The company plans cumulative investment of around 500 billion yen between fiscal 2027 and fiscal 2029. To free up that capital, Panasonic has expanded planned job cuts to as many as 12,000 positions, lifting annual savings to 145 billion yen compared with fiscal 2024. The holding company has also invested over US$10 billion in North America to acquire Blue Yonder and build out its AI platform.

What Panasonic products go into AI servers?

The Devices area contributes five product lines: data-center energy storage systems, high-performance multilayer PCB materials for GPU, ASIC and CPU circuits, conductive polymer capacitors for high-temperature environments, battery backup units for sudden load swings, and capacitor backup units for millisecond-scale power events. The company’s Award win rate for these products has reached 80 percent of the fiscal 2029 sales target.

Who is Panasonic’s partner in the AI software side?

Anthropic PBC. The partnership was unveiled at CES 2025 on January 7 as the central piece of the Panasonic Go initiative, the same day Kusumi announced the 30 percent by 2035 revenue target. Claude became the first AI assistant embedded in a Panasonic Well consumer product, the Umi wellness platform. Internally, Anthropic’s models power the PX-AI assistant used by roughly 180,000 employees and one layer of the Blue Yonder supply-chain orchestration. The two AI-software pillars now run on different clocks: the AI-infrastructure number is measured in three fiscal years, the 30 percent by 2035 target is measured across a decade.

What could break Panasonic’s AI lead?

Kusumi himself has warned that the company’s competitive advantage in storage batteries for AI data centers may not continue into the future, and that the company must be prepared to change in every direction at any time. The risk sits on both sides: hyperscalers are designing their own silicon and in-house power systems, while AI demand itself could shift toward edge or robotics deployments that need different components.