The 170 Per Cent Is Three Numbers, and the Flat One Is the Most Useful

In 2026 the Uji tencha auction turned over ¥12.94 billion, 34.4 per cent more than the year before, on 595 tonnes — the largest volume traded since 1974. Machine-harvested leaf cleared at ¥17,250 a kilogram, the highest price since the exchange began keeping records in 1996.

Divide one number by the other and something unflattering appears. Of that ¥12.94 billion, ¥10.101 billion was machine-harvested leaf: 585.6 tonnes, 98.4 per cent of everything sold. Hand-picked tencha — the tier every premium brand photographs, the stone mill, the bamboo rake, the first flush — came to 9.4 tonnes. One and a half per cent of the volume and 3.1 per cent of the money.

That is not an argument that matcha’s premium is fake. It is an argument about where the price signal actually lives, and it is the finding that reorganised our Global Matcha & Tencha Market Report when we rebuilt it this month. A market that sells its top grade at eight times the price of its bottom grade, and sells almost none of it, is not a market whose economics are set by the top grade.

Three numbers, one of which is a market

Now the figure that has been circulating: tencha prices rose 170 per cent. They did not, or at least not in any series we can find. Three exist, and they give three different answers.

Kyoto first-flush machine-harvested leaf went from ¥5,500 to ¥14,333 a kilogram — 160.6 per cent. Uji hand-picked tencha went from ¥20,024 to ¥43,330 — 116.4 per cent. And the Kyoto average for machine-harvested leaf went from ¥4,862 to ¥8,235 — 69.4 per cent. Against those, Japan’s national farm-gate average for 2025 sits at ¥8,562.

To get 170 per cent you need a 2024 base near ¥3,050 a kilogram. No published series carries one. What you have is a range of 69 to 161 per cent, which is a real and dramatic move in anyone’s currency — and the flat figure in that range is the only one broad enough to describe a market rather than a single grade or a single auction room.

This distinction is not pedantry, because the three numbers imply different businesses. A 160 per cent rise in the first-flush line says the top of the Japanese crop repriced hard. A 69 per cent rise in the Kyoto average says the general Japanese crop repriced hard. And 69 per cent is roughly what our model already contains: Japan’s leaf prices mean-reverting to USD 52.00 a kilogram by 2033 from USD 57.08 today, which is a market giving back part of a spike rather than compounding it. If you build a supply contract, a category plan or a valuation on 170 per cent, you have assumed a rate of increase that would have to keep compounding from a base that never existed.

What a price rise can and cannot fix

Here is the part that matters more than the size of the number. Tencha is not scarce because it is difficult to grow. It is scarce because there are fewer people growing it every year.

Japan produced 6,278 tonnes of tencha in 2025 from 33,400 hectares — a yield that tells you how much of that land is marginal, and how carefully the good land is worked. The grower count has fallen 77 per cent since 2000. Sixty-two per cent of core workers are over 65. A new tencha field, on the Japanese cost structure, takes about five years to pay back — which means nothing planted this season supplies a mill before 2031, and a price signal transmitted in 2026 arrives as tonnage in the next decade, if it arrives at all.

So the honest answer to a price shock in this crop is not “supply will respond”. It is “supply will respond in five to seven years, from a smaller grower base, in a country with a shrinking agricultural workforce.” Price allocates the existing crop. It does not create another one on any timetable a buyer can plan against.

A gap that does not close

That is what makes the Japan–China price differential structural rather than promotional. Japan’s 5,964 tonnes realise USD 57.08 a kilogram. China’s 11,636 tonnes realise USD 22.79. That is a gap of 2.51 times, and it exists because the two products are not interchangeable at the top and because the constraint on Japanese supply is demographic.

The Chinese side has scale that the Japanese side cannot match on any horizon. Guitea’s plant at Tongren alone holds 4,000 tonnes of capacity at one address — 34.4 per cent of the entire Chinese-origin pool and 22.7 per cent of the world pool from a single site. One company’s capacity decision is a fifth of the global market.

Which is why our model has the composition of the market changing faster than its size. Japan’s share of pool value falls from 56.2 to 35.5 per cent by 2033 while China’s share of tonnage rises from 66.1 to 77.4 per cent. Japan’s revenue still grows 20.0 per cent in absolute terms, from USD 340.4 million to USD 408.4 million. It stops being the market. It does not shrink. Those two facts are usually reported as a contradiction and are not one.

The pool, not the price

Add it up and the leaf-and-powder pool is 17,600 tonnes and USD 605.6 million in 2025, reaching 34,659 tonnes and USD 1,150.2 million by 2033. That is 8.84 per cent a year by volume and 8.35 per cent by value — and the direction of that small difference is the whole story. Realised value per kilogram falls from USD 34.40 to USD 33.19 over eight years, because the marginal tonne is Chinese and cheap. A market that grows 8.8 per cent a year in tonnage while losing 3.5 per cent a year in realised price is a volume business with a premium niche attached, not a premium business with volume attached.

You can see it more sharply in the tier structure. Ceremonial grade is 3.0 per cent of tonnage and 22.0 per cent of value — 528 tonnes at USD 252.27 a kilogram, which is 7.33 times the pool average. Premium machine-harvested is 20 per cent of tonnage at USD 65.37. Ingredient and food service is 52 per cent at USD 21.18. Industrial is 25 per cent at USD 11.02. The top tier carries a fifth of the money in a thirtieth of the volume, which means a bad harvest there is a branding event, not a supply event — and a buyer who treats it as a supply event will pay for the wrong risk.

The other structural fact is quieter. 56 per cent of tonnage is sold into someone else’s brand — food service at 30 per cent, ready-to-drink at 26 per cent. That is why a leaf-price spike shows up as a menu price or a formulation change rather than as a shelf price, and why the demand destruction a spike causes never appears in retail scanner data. It happens in formulation meetings. Anyone modelling elasticity off retail take-home (9 per cent of the pool) is measuring the smallest and most visible sliver of the demand base. The same dynamic runs through ready-to-drink and kombucha, where matcha is an ingredient in a finished product with its own price point and its own tolerance for input inflation.

Why organic is the hardest to find, and why it will stay that way

The brief that produced this piece began with two claims: prices up 170 per cent, and organic matcha the hardest grade to source. The first is a precision error. The second is exactly right, and for a reason that is worth spelling out because it is routinely confused with a price problem.

We model certified organic at 8 per cent of tonnage and 17.0 per cent of value in 2025 — roughly 1,408 tonnes at USD 73.11 a kilogram, a 2.13 times premium to the blended pool — rising to 11 and 25 per cent by 2033. Those shares are estimates, and we label them as estimates every time they appear, because we could not obtain a primary certified organic tencha area by grade from any source that publishes one.

The mechanism, though, is not an estimate. Organic supply is bounded by certified area and by conversion time, not by willingness to pay. A Japanese grower who wants to convert has to run the land through the certification cycle while absorbing the yield penalty, and the same demographic problem that is shrinking the conventional grower base is shrinking the pool of people willing to take that on. Price can reallocate the existing organic crop — it can move tonnes from a commodity blender to a premium brand, which is a very real and painful reallocation if you are the blender. It cannot create another certified hectare inside eight years. That is why organic is the hardest grade to find and why a higher price will not fix it, and it is the single most decision-relevant sentence in the whole report.

What a buyer should actually do

Three things follow, and none of them is “buy now before it goes higher”.

For procurement: the risk that matters is not price, it is access. A contract denominated in dollars against a spot index gives you exposure to a crop you may not be able to buy at any price in a bad year. Contracts that secure hectares — or a named share of a named mill’s throughput — are the ones that survive a first-flush failure. The premium for that structure is real and it is cheaper than the alternative.

For strategy and category planning: model the pool at 8.8 per cent volume and 8.4 per cent value, with realised price drifting down, and treat the 69-to-161 per cent range as a cycle, not a trend line. A plan built on 170 per cent compounding will be wrong in a direction that flatters the top line and destroys the margin.

For anyone reading the trade press: check which series a price claim came from before you repeat it. The three series in this market differ by a factor of more than two on the same year. Ours are reconcilable to their sources, and where they are not — the two official volume series over-sum by 5.7 per cent before a milling-yield adjustment — the report says so on the page rather than in a footnote.

The useful correction, in the end, is not that the price rise was smaller than reported. It is that the rise happened in a tier that carries almost no volume, in a crop whose supply is set by the age of the people growing it. Matcha’s premium is not a price story that a good season can answer. It is a land and time story that a good season can only postpone.

Figures in this article are drawn from the Global Matcha & Tencha Market Report 2026–2033. Volume is CTMA’s 17,600 tonnes reconciled against MAFF’s 6,278 tonnes of Japanese leaf at a 95 per cent milling yield; Japan is priced at the MAFF farm-gate average, China at a 70:30 blend of Zhejiang and Tongren output values. The quality tier, end use, certification and route-to-market splits are Pheonix modelled marginals, not published statistics. Related reading: herbal tea and high-protein yogurt, two categories absorbing the same premium-ingredient logic.