MIT's bacterial transistors do one calculation every eight hours. That is the useful part.

Researchers at MIT have built transistors out of bacteria, using three strains of *Pantoea agglomerans* to carry biological signals through a circuit in place of electrical ones, with the work published this week in Nature Chemical Biology ^[1][2]. Each calculation takes the cells eight hours ^[3], and that number, not the biology, is what determines where this is worth anything.

Start with the operating budget. The team reports that the combined limitations keep the living computer running for roughly three days ^[4]. At eight hours per calculation, that is on the order of nine sequential operations before the hardware expires ^[17]. Any application has to fit inside that envelope, which rules out essentially everything a conventional processor is asked to do and leaves a narrow band of problems where a three-day, nine-step computation is still early.

Christopher Voigt, head of MIT's Department of Biological Engineering and the study's senior author, told MIT News the group is "not trying to replace computers, but rather put computational control into biology" ^[5]. His example is the one that makes the timing arithmetic work: bacteria on a plant root, or the plant itself, running a simple calculation overnight, which is fast enough relative to a growth season ^[6]. The proposed use case is a circuit sitting near roots to detect stresses or respond autonomously to pests and other environmental threats ^[15]. Against a season, eight hours is not slow.

The engineering choice underneath is worth reading closely. Previous work in this area, according to the paper, mostly used enzymes to run large single cells that hosted an entire circuit, and getting a small cell to compute exactly as intended proved hard enough to cap how complex those circuits could get ^[12]. The MIT framing is that transistors "do not define circuit logic" but "conditionally enable signal propagation," so they can be recombined into more varied circuits ^[7]. In practice the team printed bacterial colonies onto small plates about 5 millimeters apart ^[8]. One molecule acts as the switch, another as a target indicating whether the switch is active, and if the target is present the transistor emits an output molecule ^[9], which then propagates to the next transistor in a chain ^[10]. Test runs produced a fair number of logical operations from five strains, with the largest circuit wiring together up to 24 colonies ^[11].

The eight hours are a consequence of that layout: signals move by natural molecular diffusion between colonies ^[13]. There is no obvious way to make diffusion across 5 millimeters fast, which means throughput is structural rather than an early-prototype artifact. The other stated limit on circuit size is that the components are alive and therefore grow and change over time ^[14].

Voigt also says that "computationally, there's nothing that your iPhone can do that these circuits couldn't do" ^[18], alongside the claim that more complicated functions can be reached by linking simpler ones in individual cells ^[16]. Read as a statement about which functions are expressible, that is a reasonable thing to say. Read as a comparison of capability in any operational sense, the three-day lifetime and the nine-calculation budget do the arguing ^[4][17].

What to watch: whether the three-day operating window extends, since it caps every application; whether the diffusion delay can be cut without abandoning the printed-colony geometry ^[13]; and whether the root-zone sensing case gets tested in soil rather than on plates ^[8][15]. Also watch whether circuits larger than 24 colonies hold their behaviour as the organisms grow ^[11][14].