Raw milk and H5N1 have been entangled since March 2024. Because the story has unfolded in fragments — a detection here, a recall there, a study that gets cited for the opposite of what it found — here is the sequence in one place, with sources.
This is an explainer assembled from public records and published research, not original reporting.
March 2024 — the spillover
Samples from affected dairies in Texas and Kansas, along with two dead cats, were confirmed as highly pathogenic avian influenza H5N1 clade 2.3.4.4b. On 25 March 2024, H5N1 in US dairy cattle was reported publicly for the first time. These viruses had not previously been found in cows.
What infected cows do to milk
The virus replicates in the udder, producing titres of up to 10^8.8 TCID50 per millilitre in milk from infected animals — an extraordinary concentration.
It also persists. In raw milk, infectious virus has a half-life of about 2.1 days at 4°C and 0.74 days at 22°C. Refrigeration, which raw milk depends on for shelf life, is close to ideal for preserving it.
July 2024 — two studies, routinely conflated
Two research letters appeared in the New England Journal of Medicine on the same day, and they are not the same study.
A NIAID-led team found the half-life of infectious virus in raw milk at 63°C was 4.5 seconds, with a 10^10-fold reduction within 2.5 minutes. A University of Wisconsin–Madison team fed mice raw milk from an infected herd and infected them; they found 63°C inactivated the virus at all intervals, while 72°C for 15–20 seconds reduced but did not always fully inactivate it.
That second result is the one that circulates as "pasteurization doesn't work." Its authors cautioned that their spiked-milk laboratory conditions do not reproduce the turbulent flow of commercial pasteurizers.
The measurement that settles it
The FDA tested 297 retail dairy samples. About 20% carried detectable influenza A RNA. None contained viable virus, confirmed by egg inoculation.
This distinction matters because the RNA finding is itself frequently reported as alarming. Genetic fragments of an inactivated virus are precisely what you would expect to find downstream of pasteurization that worked.
November–December 2024 — California
24 November: California detected H5N1 in a retail sample of Raw Farm cream-top whole milk, prompting a voluntary recall of one lot. 27 November: a second lot. 3 December: a broad recall of all Raw Farm raw whole milk and cream produced 9–27 November, after the virus was found at both retail and the dairy's own storage and bottling sites. The California Department of Food and Agriculture placed the farm under quarantine.
No human cases were linked to the recalled product.
Where things stand
The CDC counts 71 human A(H5) cases in the United States since February 2024, with two deaths — overwhelmingly among people with direct exposure to infected dairy herds or poultry. There is no known person-to-person transmission, and the assessed public health risk remains low.
In December 2025, a Wisconsin dairy herd detection was sequenced as genotype D1.1 — a new spillover from wildlife, distinct from the B3.13 genotype behind most dairy cattle cases. The route from wild birds into cattle is not closed.
A number we are not going to give you
You will see it reported that raw milk sales surged 65% after the outbreak news. What the source actually says is that weekly sales rose "21% to as much as 65%" year-over-year in May 2024 — a range, reported as its own maximum.
It is also a range on a very small base. That same reporting notes roughly 4,100 units of raw cow's milk sold in a week against about 66.5 million units of pasteurized milk. And it is from May 2024, which is before the recalls, though it gets recycled as though it were current.
There may well be a real story about defiance driving raw milk consumption. This statistic is too weak to carry it.