From ENIAC to the Law of Tau: what if computing changed its compass?

Reading note: this article discusses an idea attributed to Huawei, sometimes

called the "Law of Tau." The concept circulates more than it is documented. We

therefore treat it as a proposal to examine, not as an established fact.

In one sentence

For sixty years, the industry measured computing progress in space — how many transistors on a chip; the proposal we examine here suggests measuring it in time — how fast information travels, which would be less a rupture than a return to the real story of the field: efficiency.

1. Setting the scene

In 1945, ENIAC enters service. Thirty tons. Nearly 18,000 vacuum tubes. Power consumption comparable to that of a city block. At the time, the machine performed in seconds what had previously taken hours of human calculation.

The phone in your pocket now exceeds that power by several orders of magnitude. This trajectory is often summed up in a single phrase: Moore's Law.

We write from Réunion Island, 9,000 km from the leading-edge foundries. From here, one question keeps coming back: when a technology roadmap stops working for some players, what do they replace it with? Huawei, constrained by US sanctions, seems to offer an interesting answer — not "do the same thing, but worse," but "change the quantity you are trying to optimize."

It is this shift, more than the technical feat, that interests us.

2. Key definitions

Moore's Law is not a physical law. It is an observation, formulated by Gordon Moore in 1965: the number of transistors integrated on a chip doubles at regular intervals. For decades, it served as a compass for the entire industry.

The "Law of Tau" names a different idea. The symbol τ (tau) represents, in electronics, a time constant: the characteristic duration a signal takes to settle in a circuit. The lower τ is, the faster information travels. The idea would be to make this travel time — rather than component size — the reference measure of progress.

Let us be clear right away: to our knowledge, "Law of Tau" is not a theoretical framework formalized and validated by the industry in the same way as Moore's Law. It is, at best, a framing proposal. We discuss it as such.

3. Analysis

3.1 Miniaturization was never the story

Computing is often told as a race toward smallness. From vacuum tubes to transistors, from integrated circuits to microprocessors, each step shrinks the components.

But shrinking has never been an end in itself. ENIAC was not surpassed because it was big. It was surpassed because it was inefficient — in energy, in heat, in how it organized information. Miniaturization was long the simplest way to gain efficiency. It was not the goal.

That nuance matters, because it changes how we read what is happening today.

3.2 The bottleneck has moved

For a long time, the scarce resource was compute. That is no longer quite the case. Modern processors have it in abundance, and specialized AI accelerators add even more.

The brake now lies elsewhere: in the movement of data. Training and running a large model means constantly shuttling information between memory, compute units, accelerators, and the network. Each trip takes time and consumes energy.

This observation is not specific to Huawei; it is widely shared in the hardware community. The problem is sometimes summed up like this: in some systems, moving a piece of data would cost more than computing it. We present this as an often-cited order of magnitude, not a universal measurement — the exact ratio depends on the architecture and the workload.

3.3 From the reign of space to the reign of time

If we accept this diagnosis, the idea behind the "Law of Tau" becomes legible. Rather than asking "how many transistors per unit of area?", it asks "how long does a piece of information take to reach its destination?".

The shift seems subtle. Yet it orients design differently: the aim is less to etch finer than to shorten the distances data must travel.

And this is no isolated whim. Several already well-established trends point the same way:

  • chiplets, which split a chip into blocks connected as closely as possible;
  • 3D stacking, which layers components to bring memory and compute together;
  • high-bandwidth memory (HBM), placed as close as possible to accelerators;
  • specialized architectures for AI, tailored to reduce round trips.

All pursue the same goal: bringing data closer to compute. Huawei's proposal consists, in part, of naming what is already underway — and making it an explicit compass rather than a side effect.

3.4 A constraint turned into an angle

It would be naive to ignore the context. Cut off from access to the most advanced etching processes, Huawei has a direct interest in shifting the terrain of competition. If you can no longer win the nanometer race, better to redefine what counts as a victory.

This does not invalidate the idea. A good part of the history of innovation consists precisely of turning a constraint into an angle. But it calls for caution: we do not yet know whether the "Law of Tau" is a robust framework set to structure the industry, or a clever narrative serving a workaround strategy. Both readings coexist, and the future will decide.

4. Implications

If this framing takes hold, several things move.

First, the hierarchy of skills. Optimizing travel time rewards architecture, packaging, and system integration as much as — if not more than — etching fineness. And those skills are less exclusively concentrated in a single foundry.

Second, the measure of performance. As long as the world compares nanometers, a player without access to the latest processes is structurally outpaced. If the conversation shifts toward energy efficiency and latency, the rankings reopen — at least in part.

Third, the economics of data centers. In the AI era, energy has become a central cost line. Any framework that puts efficiency first, rather than raw power alone, speaks directly to the operators who pay the electricity bill.

We stay measured: none of these implications assumes that Huawei "replaces" Intel, Nvidia, or TSMC. The real question is not a corporate duel, but a possible change of compass for the sector as a whole.

5. Signals to watch

  • Formalization. Does the "Law of Tau" lead to publications, reproducible metrics, benchmarks — or does it remain a communications element?
  • External uptake. Do players other than Huawei explicitly adopt travel time as a reference indicator?
  • Concrete products. Do chips claiming a measurable gain in latency and energy, at constant etching fineness, appear on the market?
  • Benchmark vocabulary. Do public comparisons keep speaking in nanometers, or do they shift toward "useful work per watt"?
  • Standardization. Do consortia or standards bodies take up the notion?

6. A situated word

Seen from Réunion, this story resonates beyond hardware. On a remote island, where electricity is expensive and the grid imposes its limits, efficiency is not a slogan: it is a condition of existence. Doing more with less is not an aesthetic choice, it is daily life.

That may be why the idea speaks to us. If the industry stops conflating progress with excess and relearns to measure what is truly lost — time, energy, useless trips — then it joins an intuition that constrained territories have known for a long time.

ENIAC was the age of brute force. Moore's Law, the age of miniaturization. The "Law of Tau," if it keeps its promises, could open the age of deliberate efficiency. As often in technology, the real turning points are not about going bigger. They are about doing better.

We do not yet know whether this is one of them. We find the question fair enough to be worth asking.

Sources and further reading

Note: as of publication, the "Law of Tau" attributed to Huawei has no formalized primary source we can cite; we treat it as a framing proposal to verify, not as an established framework.

Written from Réunion Island, where it is already tomorrow.