Samsung Electronics guided to a record $80bn result after flagging a 783% year-on-year jump in profits, with the Korean chipmaker attributing the swing to an AI semiconductor cycle that is also lifting sales at TSMC, according to a Yahoo Finance UK report on the company's earnings disclosure.

The 783% figure measures profit against a depressed prior-year base, but the absolute $80bn mark is the headline number Samsung is pointing to as evidence that memory pricing has turned and high-bandwidth memory (HBM) shipments for AI accelerators have scaled faster than the company's own internal forecasts had assumed.

What is driving the Samsung numbers?

• Memory price recovery after the 2023 downturn
• HBM volume tied to Nvidia-class AI accelerator demand
• Foundry utilization gains on advanced nodes
• A weak year-ago comparable that amplifies the percentage move

Samsung has been chasing the HBM3 and HBM3E sockets that Nvidia's H100 and H200 GPUs rely on, while also qualifying for next-generation accelerators. Each shipped HBM stack carries materially higher revenue density than the DDR5 and LPDDR lines that dominated the 2023 mix.

How is TSMC caught in the same tailwind?

TSMC, the world's largest foundry by revenue, separately confirmed that AI chip demand is lifting its top line. The Taiwan-based manufacturer is the sole source of Nvidia's most advanced GPUs on its 5nm and 4nm processes and is ramping 3nm for Apple's A-series and M-series silicon as well as AMD's MI300 family. Capacity at TSMC's Fab 18 in Tainan and Fab 21 in Arizona remains the gating factor for AI accelerator supply globally.

Is this a one-quarter effect or a structural shift?

Samsung framed the print as a record, suggesting management views the level rather than just the year-on-year percentage as the relevant benchmark. Two structural drivers support that reading:

• AI training and inference capex at hyperscalers (Microsoft, Google, Meta, Amazon) continues to accelerate, and each accelerator pulls multiple HBM stacks and a large die fabricated on a leading-edge node.
• Traditional memory end markets — smartphones, servers, PCs — have also stopped bleeding inventory, lifting commodity DRAM and NAND pricing after four straight down quarters.

What does this mean for foundry and memory pricing?

If Samsung's $80bn print holds in the audited results and TSMC's AI-linked revenue mix continues to widen, suppliers can expect:

• HBM contract pricing to remain firm through at least the first half of next year
• Foundry leading-edge wafer prices to stay elevated as TSMC allocates capacity by customer weight
• Competitive pressure on Samsung's foundry division to close its yield gap with TSMC on 3nm and to translate its HBM momentum into advanced-packaging wins

The risk to the picture is concentration: both Samsung's memory surge and TSMC's AI lift depend on a small number of hyperscaler and accelerator customers whose purchasing patterns are visible only through indirect signals such as Samsung's preliminary filings.

Samsung's full earnings release will pin down whether the $80bn figure represents operating profit, gross profit, or a top-line metric, and whether the 783% year-on-year move includes any one-off items such as chip inventory write-back reversals. Until then, the preliminary disclosure sets the bar: AI demand has pushed the Korean group to a record print, and TSMC is reading the same demand signal in its own order book.