The standard industrial method for making custom DNA relies on hazardous organic solvents. A Harvard team has replaced that chemistry with a water based enzymatic reaction wired directly to a silicon chip.
The standard industrial method for making custom DNA strands relies on toxic organic solvents flowing through bench-sized synthesizers in centralized labs. A Harvard team has now run that same job on a silicon chip the size of a thumbnail, swapping the solvents for water and the plumbing for electricity.
In a paper published this week in Nature Electronics, the same group, led by Donhee Ham at the Harvard John A. Paulson School of Engineering, reports a complementary metal-oxide-semiconductor (CMOS) chip that synthesizes 64 different DNA sequences in parallel on a single surface. Each strand is assembled by an enzyme called terminal deoxynucleotidyl transferase, or TdT, a template-independent polymerase that, in nature, helps immune cells diversify their antibodies. The lab's innovation is wiring TdT to an electrode grid, so the chip's electrical signals pick which nucleotide lands at which spot, replacing the wash-and-deprotection cycles of conventional phosphoramidite chemistry with a controlled aqueous reaction.
The Harvard SEAS press release frames the work as an alternative manufacturing route rather than a finished product. Independent re-coverage by ScienceDaily, Phys.org, and Bioengineer.org all converge on the 64-strand parallel figure.
The mechanism matters because it borrows the semiconductor playbook. Conventional DNA synthesizers treat strands one well at a time, in a sequence of reaction chambers fed by tubes and valves. The CMOS approach treats the chip surface itself as an addressable array. Each electrode is a reaction site; each current pulse is a reagent delivery step. In principle the same approach scales the way memory and logic chips scale, by shrinking the cell and adding more of them to the wafer. That parallelism is the most plausible path to bringing the cost of writing a custom gene down by orders of magnitude. The same cost curve turned computing from a room-sized enterprise into a phone in every pocket.
DNA has been pitched for years as an archival storage medium, with theoretical densities far beyond magnetic tape. The bottleneck has been writing speed and cost, both of which trace back to the same solvent-heavy chemistry the Ham lab is now trying to retire. A water-based, electrically addressed synthesizer would also make custom DNA manufacturing portable, replacing benchtop robots the size of refrigerators with instruments a single graduate student could carry.
The demonstrated sequences are short, and the paper itself notes that scaling strand length, yield, and fidelity will require new chemistry beyond the TdT reaction shown. The chip is a working prototype; the 64 sequences are not yet at the length or error rate of commercial phosphoramidite output.
That caveat is the gap between a research milestone and a manufacturing platform. The Ham lab is publishing the proof of concept first; the chemistry that would let the chip write genes thousands of bases long at the fidelity biologists expect is the next paper, not this one.
The work reframes how custom DNA could be manufactured at scale. For forty years, custom DNA has been the product of large, centralized foundries running solvent-intensive chemistry. The Nature Electronics paper proposes a different path, treating DNA synthesis as a form of lithography with the silicon industry as its fab. Whether the chemistry scales is the next thing to watch.