Samsung Makes Almost Everything in Its Own Phone. In 2023 That Cost It KRW 14.88 Trillion.

Crack open a Samsung Galaxy and you are holding a small civil war that resolved in Samsung's favor. The processor, the DRAM, the NAND flash that stores your photos, the screen you stare at - in a Samsung phone, all four can come from Samsung itself. Those four parts make up roughly two-thirds of a phone's bill of materials.⁵ No other phone maker comes close to that. Apple designs its chips but has them built by someone else; almost everyone else simply buys. Samsung is, uniquely, both the store and the factory behind it.

The official story is that this is the secret to Samsung's dominance: own the whole stack, capture every margin, never wait on a supplier, out-cost everyone. It is a beautiful story. In 2023 it produced a record loss of KRW 14.88 trillion in the very division that was supposed to be the engine.³ The integration didn't fail despite being complete. It hurt because it was complete.

What owning the whole phone actually buys you

Start with what the integration genuinely delivers, because it is real. When you build your own processors, memory, and display, you don't negotiate for supply - you allocate it. You set your own technology roadmap instead of waiting for a vendor's. You ship a new screen in your flagship before any competitor can buy the same panel, because you own the panel maker. The Device Solutions division alone booked KRW 66.6 trillion of revenue in 2023, more than a quarter of the company.¹ Samsung's display arm added another KRW 31 trillion.¹ These are not internal cost centers; they are some of the largest component businesses on earth, and they happen to sit under the same roof as the phone that needs them.

But here is where the tidy story gets murky. The popular claim is that making it yourself guarantees a cost advantage. It doesn't, and the people closest to the numbers say so. Samsung's divisions sell to each other at arm's-length transfer prices - the phone unit pays roughly what an outside buyer would. One analysis pegged the parts savings at a meaningful slice of the bill of materials; another, in the same breath, questioned whether any net advantage exists at all once you account for those internal prices.⁵ Integration buys control. Whether it buys cheapness is genuinely contested.

Why a glut hits the integrated player twice

Now the thesis. Vertical integration is a leverage machine, and leverage works in both directions. In a stable commodity market, owning the supply chain compounds quietly in your favor. In a downturn, the same plumbing carries the pain straight through to your bottom line at full pressure. 2023 was the demonstration. A post-pandemic inventory glut sent memory prices into free fall - roughly 70% over nine months - and because Samsung is one of the biggest memory makers in the world, it ate the collapse in proportion to its scale.⁴ By Q1 2023 its quarterly operating profit had cratered 96%, its weakest quarter since the depths of 2009, and it took the rare step of announcing a 'meaningful' production cut to stop bleeding the price further.⁴

Read the full-year ledger and the asymmetry is stark. The company stayed profitable overall - KRW 6.57 trillion in operating profit for 2023 - but only because the consumer and mobile side carried it while semiconductors hemorrhaged.² The division everyone calls Samsung's profit engine was, that year, its largest drag. And notice the capital trap underneath: Samsung poured KRW 48.4 trillion of capex into Device Solutions in 2023 even as it lost money there.² You cannot idle a fab the way you can idle a purchase order. When you own the factory, the downturn finds you whether you produce or not.

When being the maker poisons your customers' trust

There is a second, subtler cost, and it shows up exactly where Samsung most wants to grow: the contract chip business. Samsung Foundry wants to manufacture chips for other companies, the way TSMC does. But Samsung is also those companies' competitor - it designs its own processors and sells its own phones. Handing your blueprint to a foundry owned by your rival is a hard sell, and the market reflects it. Samsung Foundry sits around 11-13% of the global foundry market against TSMC's roughly 62%.⁷ The neutrality TSMC sells - 'we build, we never compete' - is precisely the thing Samsung's integration forbids it to offer.

The most telling detail is in Samsung's own filing. Its 2023 Business Report lists Qualcomm as both a top customer and a key supplier.¹ Qualcomm's Snapdragon chips have powered Galaxy flagships in recent cycles - chips fabricated by TSMC, not by Samsung's own foundry. The supposedly self-contained company buys its best processors from a partner who builds them at the rival next door. And the integration is fraying further at the edges: trade reports say Samsung has even weighed outsourcing some advanced memory-chip base-die production to TSMC, with its own cutting-edge 2nm yields reported well below commercial thresholds.⁷ The maker, in its hardest market, is quietly becoming a buyer.

Isn't this just one bad year for an enduring model?

The fair objection is that 2023 was a cyclical trough, not a structural verdict - memory has always boomed and busted, and Samsung has survived every cycle by spending through them. There's truth in that. Samsung's R&D was KRW 28.4 trillion in 2023, near 11% of revenue, and rose again in 2024.⁸ No company outspends downturns like Samsung, and integration is exactly what lets it keep its component roadmaps moving while rivals stall. In the next memory upcycle, the same leverage that gutted 2023 will swing hard the other way. The model isn't broken; it's volatile, and volatility cuts both ways.

But the honest counter doesn't fully rescue the cost-advantage myth. The cyclical argument concedes the point - integration amplifies the cycle rather than smoothing it, which is the opposite of the 'protection' story. And the foundry problem isn't cyclical at all; it's structural and getting worse. You can spend your way back into a memory boom. You cannot spend your way out of being your customers' competitor. Apple, meanwhile, overtook Samsung as the world's No. 1 smartphone vendor in 2023 with a 20% share - and it did so owning almost no fabs at all.⁶ The lightest stack beat the deepest one in the same year the deepest one lost a fortune.

Samsung built the most complete machine in consumer electronics - a company that can make nearly everything in the device it sells. That completeness is genuine power in stable commodity markets, and a genuine liability everywhere it has to win someone else's trust. The integration was never a guarantee of low cost or a shield against the cycle. It was a bet that owning everything beats buying anything. In 2023 the bet came due, all at once, in a single line on the ledger - and proved that when you own the whole stack, you also own the whole fall.