Samsung Projects Record $80 Billion Annual Profit as AI Chip Demand Reshapes Its Business

Today’s Focus

Samsung Electronics said Wednesday that its third-quarter operating profit rose about ninefold from a year earlier, driven by booming demand for the high-bandwidth memory (HBM) chips that power generative artificial intelligence systems, according to the Financial Times.

The company’s guidance put quarterly operating profit at roughly 12.1 trillion won, and analysts cited by Yahoo Finance now expect full-year operating profit to approach $80 billion, which would be a record for the Suwon-based conglomerate.

The surge comes after more than a year in which Samsung lagged rival SK Hynix in supplying HBM chips to Nvidia, the dominant designer of AI accelerators. Samsung said in its guidance that memory revenue climbed sharply as hyperscale cloud customers accelerated orders.

TradingView reported that Micron, a U.S. memory rival, lifted its own revenue outlook by roughly $70 billion over the coming cycle, a sign that the AI-driven memory shortage is industry-wide rather than company-specific.

The windfall is reshaping Samsung’s product mix. Digital Trends, citing supply-chain reports, said Samsung is producing roughly 30% fewer smartphones this year because handset margins no longer justify the capacity, with wafer allocations shifting toward HBM and foundry orders.

The Investor, a Korean trade outlet, reported that Samsung’s mobile division is paying higher internal prices for the same memory chips its sister division sells to Nvidia and Google, squeezing Galaxy profitability even as group earnings soar.

Samsung will release detailed divisional results later in October. The preliminary guidance is a legally required disclosure under South Korean securities rules and does not break out profit by business line.

The Debate

Supporters argue

Backers of Samsung’s pivot say the numbers vindicate a years-long bet on advanced memory. Counterpoint Research analyst MS Hwang told the Financial Times that Samsung has “finally caught up” in HBM qualification with Nvidia and is now positioned to capture share through 2027.

South Korean President Lee Jae-myung’s office has framed the semiconductor boom as proof that Seoul’s industrial policy, including tax credits for chip fabrication passed in 2025, is paying off. Trade Minister Ahn Duk-geun said in a statement cited by Reuters that memory exports are driving the country’s trade surplus to multi-year highs.

Industry supporters also argue the smartphone cutback is rational. Wedbush analyst Dan Ives, quoted by Yahoo Finance, said the AI capital-spending cycle is “a generational demand wave” and that reallocating fabrication capacity toward data-center chips is the correct response for any integrated device maker.

Samsung executives told investors on prior calls that HBM4, the next memory generation, is on track for mass production in early 2027, which bulls say should extend the earnings run.

Critics argue

Skeptics warn Samsung is becoming dangerously dependent on a single customer base. Sanford C. Bernstein analyst Mark Li, cited by The Investor, said concentration risk is rising because “a handful of hyperscalers” account for the bulk of HBM orders, leaving Samsung exposed if AI capital spending slows.

Consumer-electronics critics say the smartphone cuts hollow out a franchise Samsung spent two decades building. Digital Trends noted that reduced Galaxy output cedes shelf space to Chinese rivals Xiaomi and Transsion in emerging markets, where switching costs are low.

Opposition lawmakers in Seoul have questioned whether the chip boom is widening inequality. Democratic Party lawmaker Kim Byung-kee told local media that record corporate profits have not translated into wage gains for subcontracted workers at Samsung’s Pyeongtaek complex.

Some financial analysts also flag valuation risk. Macquarie’s Daniel Kim wrote in a note cited by Yahoo Finance that Samsung shares already price in sustained HBM pricing power, and any crack in AI demand “would be felt first in Suwon.”

What the experts say

Independent researchers describe the current cycle as the sharpest memory upswing since the 2017-2018 boom, but with a narrower demand base. The Semiconductor Industry Association reported global chip sales reached $627 billion in 2024 and projected further double-digit growth through 2026, with AI accelerators and the memory that feeds them accounting for most of the increase.

Gartner analyst Gaurav Gupta told Bloomberg this summer that HBM revenue could more than double in 2026 to above $50 billion, which would make it the single most profitable memory segment in history.

The International Energy Agency’s 2024 report on electricity and AI estimated data-center power demand could double by 2030, a trajectory that implies continued pressure on chip supply unless efficiency gains outpace model scaling.

Historical context cuts both ways. The U.S. Bureau of Labor Statistics has documented three memory-industry downturns since 2000 in which prices fell by more than 40% within 18 months of a peak. Economists at the Korea Institute for International Economic Policy have warned that past cycles ended abruptly once hyperscaler capital budgets normalized.

By the Numbers

12.1 trillion won: Samsung’s preliminary third-quarter 2026 operating profit, roughly nine times the figure from the same quarter a year earlier, according to the company’s regulatory filing cited by the Financial Times.

$80 billion: approximate 2026 full-year operating profit now forecast by analysts tracked by Yahoo Finance, which would be a company record.

30%: reported reduction in Samsung smartphone production this year as capacity shifts to memory and foundry, per Digital Trends.

$70 billion: upward revision to Micron’s multi-year revenue outlook cited by TradingView, reflecting industry-wide AI memory tightness.

$627 billion: global semiconductor sales in 2024, per the Semiconductor Industry Association.

$50 billion+: projected 2026 HBM revenue, according to Gartner analyst Gaurav Gupta as reported by Bloomberg.

40%+: typical peak-to-trough memory price decline in the three downturns since 2000 tracked by the U.S. Bureau of Labor Statistics.