On Tuesday China’s customs administration published its trade figures for August, and at first glance they read like a declaration of war. Exports rose 25 percent to $401.4 billion, the trade surplus reached $119.1 billion in a single month, and the cumulative surplus for the first eight months stands at $805.5 billion, on track to beat last year’s record of $1.19 trillion. Shipments to the United States jumped 34.4 percent, lifting the bilateral surplus with America to $29.2 billion, the widest since Donald Trump returned to the White House. In eleven days, on September 24, Xi Jinping arrives in Washington for a state visit with a large business delegation in tow. It does not take a prophet to guess which number will be on the table.
It will be the wrong number. Because anyone who reads one line further down in the same release finds a fact that turns the picture upside down: the value of exported semiconductors rose 129.8 percent, while the number of units exported fell 8 percent. The value of exported data-processing equipment, meaning servers and computers, grew 76.5 percent while volumes declined by double digits. China did not ship more chips to the world in August than a year earlier. It shipped fewer, and collected more than twice as much for them. The record surplus that Washington and Brussels are arguing about is, in its fastest-growing component, not a volume phenomenon but a price tag. And price tags, which is the point of this commentary, are not negotiated at summits. They change on their own.
Value versus volume: the line beneath the headline
For its main product categories, Chinese customs publishes two series, value and units, and normally they move in the same direction. In August 2026 they did not. Integrated circuits brought in $40.7 billion of export revenue, against roughly $17.7 billion in August 2025. That is an increase of $23 billion in one month, on a unit count 8 percent below the prior year. Run the arithmetic and the average revenue per exported chip has risen by a factor of about 2.5 in twelve months. For the first eight months of the year the agency reports chip exports of $256.8 billion, up 103.9 percent; in yuan terms it is 1.77 trillion and 95.4 percent, the gap being the currency’s appreciation against the dollar.
Where the price jump comes from is no secret, and readers of this column have seen it several times: memory chips, DRAM and NAND, have multiplied in price since early 2025 because the three big producers have redirected their wafer capacity toward accelerators for data centers. China is not the place in this chain where memory is made. It is the place where it is packaged, tested, soldered onto modules and built into servers. The import side of the same release shows exactly that: imports from South Korea rose 108.1 percent in August, imports from Taiwan 41.5 percent, imports of data-processing equipment 209 percent, chip imports over eight months 61.7 percent. China buys the components that have become expensive in Seoul and Hsinchu, adds a layer of work and sells them on. The markup that accrues along the way is booked on both sides of the border, once as an import and once as an export.
The aggregate confirms it. Exports were $80.3 billion above the year-earlier month, imports $62.1 billion above. The surplus therefore grew by $18.1 billion, from $101.0 billion to $119.1 billion. Chip revenue alone rose by $23 billion. Put differently, the entire increase in the surplus can be explained by the price rise in a single product group, and one whose inputs China itself imports at sharply higher prices. Of the $80 billion in additional exports, $18 billion stayed in the country. The rest is pass-through.
Where the surplus arises, and where it does not
The regional breakdown says the same. Shipments to the United States reached $42.5 billion, up 34.4 percent after 17 percent in July. That sounds like an export offensive against the tariffs, but it is mostly a base effect: in August 2025, after the spring tariff increases, exports to America had collapsed by roughly a third, and even after this jump they remain below the level of 2024. The US share of Chinese exports, as Capital Economics notes, has stabilized at around 10 percent. The real shift is happening within Asia: exports to Southeast Asia rose 30.2 percent, exports to South Korea and Taiwan by more than 40 percent. Those are the countries from which China sources its chips and to which it sends modules and equipment back. Trade inside the supply chain is inflating with the prices, in both directions.
Europe barely appears in this picture. Exports to the European Union grew 6.6 percent, the weakest reading in ten months. Imports from the Union grew 0.7 percent. Both numbers matter, for different reasons. The 6.6 percent means the European market is not participating in the AI trade, neither as a buyer of servers nor as a supplier of chips. The 0.7 percent means European exporters to China, above all German machinery makers, chemical groups and carmakers, are getting nothing from the Chinese import wave. China imported 28.2 percent more in August than a year earlier, 44.6 percent more from Australia, 20 percent more from Japan, 17.8 percent more from the United States. From Europe, practically nothing. That is the number Brussels ought to be discussing with Beijing, not the surplus.
The deflation exporter that no longer is one
A day after the trade figures came the second statistic that belongs to this picture. China’s producer prices rose 3.8 percent year on year in August, after 3.5 percent in July and above the 3.6 to 3.7 percent economists expected. On its own that is an unremarkable number. In context it is historic: from October 2022 through February 2026, Chinese factory prices fell every single month, the longest deflationary stretch in decades. Only in March of this year did the index turn positive, at 0.5 percent, and it has accelerated since. The economy that supplied the world with ever-cheaper goods for three years is now supplying them at higher prices.
Here too it pays to look beneath the headline, because the 3.8 percent is unevenly distributed. Mining prices rose 17.8 percent, raw materials 6.7 percent, processing 3.1 percent. Consumer goods prices, by contrast, kept falling, by 0.5 percent, with food down 2.3 percent and clothing down 1.2 percent. Only durable consumer goods, meaning appliances, electronics and furniture, flipped sign: up 1.2 percent after 0.4 percent in July. That is the first place where the memory price shows up in China’s domestic statistics, because every television and every laptop contains a module that has become more expensive. What China ships to Europe and America is to a large degree exactly this category. The inflation impulse that the Fed is likely to answer with a rate increase on Wednesday no longer comes only from oil out of the Strait of Hormuz; it also comes from a supply chain that for three years delivered the opposite.
What Beijing itself says
It is worth noticing how the Chinese side presents the figures. In its statement, the customs administration emphasizes not the surplus but imports: they grew faster than exports for the sixth consecutive month, in yuan terms 21.7 percent against 18.6 percent in August. Over eight months imports rose 22 percent, exports 14.6 percent. That is not cosmetics; it is the flip side of the same price movement. Whoever imports the components that have become expensive necessarily reports rising imports. The cumulative surplus of $805.5 billion is therefore only a few percent above the same period last year, even though total trade volume grew 17.6 percent. The surplus is growing more slowly than trade. That is exactly what one would expect when a price wave passes through a supply chain with the same country at its beginning and its end.
The domestic economy, for which exports are supposed to be the lifeline, looks no better meanwhile. The manufacturing purchasing managers’ index contracted for a second straight month in August, consumer prices rose 0.8 percent, food prices are falling. On the Sunday before the release, the government announced it was injecting around $54 billion into state banks and insurers. Alicia García-Herrero of Natixis reads the import figures as corporate spending on computing capacity and electronics, not as a recovery in household consumption. Zhiwei Zhang of Pinpoint puts it more bluntly: China continues to rely on the exporters to support the economy. Wen Bin of Minsheng Bank expects the AI-driven capital-expenditure cycle to keep supporting both volumes and prices across the semiconductor chain. Note the order: volumes and prices. Volumes are already falling.
The summit and the wrong number
The political calendar is tight. At the G20 finance ministers’ meeting in Asheville, 19 of 20 members agreed to address trade imbalances; China was the lone dissenter, after Treasury Secretary Scott Bessent had described the Chinese surplus as a barrier to global growth. The European Union, which by its own count runs a trade deficit with China of roughly one billion euros a day, introduced safeguards for its steel industry in July, curtailed duty-free imports of small parcels from Chinese e-commerce platforms and scheduled ministerial-level trade talks for the fall. And in Washington, from September 24, Xi Jinping will sit across from a president who reads trade balances as a scoreboard.
The problem with this constellation is that the number under discussion will be a different one in twelve months, regardless of what is agreed. If memory prices normalize, and contract-price curves have pointed toward a flattening since the third quarter, then the value of China’s chip exports, the value of its chip imports and the surplus will all fall together, without a single container fewer sailing. An agreement that reduces the surplus will then be celebrated as a success that nobody earned. Conversely, should the price wave persist, no agreement will stop it. Chi Lo of BNP Paribas Asset Management describes the situation as a strategic stalemate in which each side holds the other hostage with products, the United States with high-end technology, China with rare earths. Both hostages are quantities. The surplus is a price. One negotiates the former and measures the latter.
What it means for US investors
The obvious question is who profits from a trade boom made of prices. The answer is different from what reflex suggests. Anyone who counts containers does not see this boom. FedEx, which reports on Thursday, UPS, Maersk and the Pacific carriers are paid by volume, by boxes, pallets and tons, and the unit counts of the most valuable product category are declining. A chip that costs twice as much takes the same space on a ship. Transpacific freight rates reflect that: they have nothing in common with the export values. Reading the August figures as a buy signal for logistics is reading the statistics wrong.
On the other side of the ledger sit the American buyers of what China assembles. Dell, Hewlett Packard Enterprise and Super Micro import the servers and modules whose values rose 76 and 130 percent; their revenue lines swell with the pass-through, their gross margins do not, and both have already announced double-digit price increases to customers. Apple, as we wrote on Thursday, absorbs part of the memory cost in its iPhone pricing. The only place where the price wave is captured as profit is where it originates, at Micron, Samsung and SK Hynix, with every caveat about cycle timing that we have laid out in recent weeks. Micron’s single-digit forward earnings multiple is the market’s way of saying that it expects the pass-through to reverse.
There is a policy angle as well. The bilateral surplus of $29.2 billion is the number the administration will cite; the base effect that produced it will expire in September, when Chinese exports to the US start comparing against months in which tariffs already applied. Growth rates will fall by arithmetic, and it would be a mistake to read that as a summit outcome. Semiconductor tariffs under Section 232, should they be extended to assembled modules, would raise the price of exactly the imports whose value is already inflated. For taxable accounts, gains on any of these positions held over a year are taxed at long-term capital gains rates of up to 20 percent plus the 3.8 percent net investment income tax; positions in Korean and Taiwanese names carry foreign withholding on dividends that requires more paperwork to reclaim than a W-8BEN.
Risks and counterarguments
The thesis of this commentary, that the surplus is a price phenomenon, can be attacked at several points, and one should know them. First, not all exports are semiconductors. Auto exports rose 43 percent, ship exports 21 percent, and labor-intensive goods such as toys rebounded in August after falling 0.6 percent over eight months. Those are volume gains, and they are real. They are not new, however; they have driven European trade complaints for two years, and they do not explain the surplus jump this summer.
Second, the base effect on US-bound exports is expiring, as noted. Third, the price wave could last longer than the forward curve implies. If the memory makers push their allocation further toward accelerators in 2027, module prices stay high and Chinese export values stay inflated. The surplus would still be a price tag, but a durable one. Fourth, and this is the most serious objection: the rise in producer prices could be a regime change rather than a commodity effect. If Chinese industry regains pricing power after three years of deflation because the government is actually cutting overcapacity, then the country will export inflation structurally from here on. Consumer goods prices, still falling, argue against that so far. The turn in durable goods argues for it.
Outlook: three numbers that say more than the surplus
The summit on September 24 will produce a number, probably an extension of last October’s trade truce, perhaps purchase commitments for soybeans, whose imports fell 1.1 percent in August, or for aircraft. For investors, three other numbers matter more. The first is the customs unit series for semiconductors and data-processing equipment; the next release on October 14 will show whether volumes keep falling while values rise. As long as that is the case, the surplus is a price. The second is the consumer goods line in the producer price index: if it stays negative, China remains a deflation anchor for Western consumer prices, and the central banks are fighting oil and memory with their rate increases, not Beijing. If it turns positive, the situation for the Fed and the ECB changes fundamentally. The third is the 0.7 percent, import growth from Europe. It is the only number in this release that a negotiation could actually change, and the only one nobody is negotiating.
What remains is a record surplus that consists in large part of a price movement China neither triggered nor controls, that is read in Washington and Brussels as evidence of an export offensive which does not exist in the unit counts, and that will shrink on its own when memory prices turn, at a moment when everyone involved will claim it as a negotiating success. The deflation China delivered to the world is over. The surplus the world is arguing about is its last echo: it measures what the world pays for the building blocks of its data centers, and it measures it at the border of a country that merely passes them along.
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.
