Harvard's DNA-writing silicon chip: how it works and why it matters

What the Harvard team actually built

Researchers at Harvard’s John A. Paulson School of Engineering and Applied Sciences, led by electrical engineer Donhee Ham, report a silicon chip that synthesizes 64 different DNA strands simultaneously. The result was published in Nature Electronics and summarized by ScienceDaily. Rather than using the phosphoramidite chemistry that has dominated synthetic DNA production for decades, the device drives DNA building with water-based enzymatic reactions triggered by precise electrical currents on the chip surface.

Each of the chip’s 64 synthesis sites carries a pair of concentric ring electrodes surrounding a single DNA molecule anchored to the surface. When a site is activated, the inner electrode produces protons, lowering the local pH just enough to remove a temporary blocking group from the growing strand—a step called deprotection—so the next nucleotide can be attached. The outer ring electrode simultaneously neutralizes protons that drift outward, confining the acidic zone to that one site. Repeating this across many cycles builds a unique sequence at each location.

The longest strand the team reported was 39 nucleotides, and the device successfully synthesized 64 sequences in parallel. The researchers describe this as a step up from earlier enzymatic-demonstrations, which ScienceDaily says were limited to roughly a dozen sequences at once.

Why this is different from how DNA is normally made

Almost all custom DNA sold to research labs and biotech companies today is produced by phosphoramidite chemistry, an established method capable of manufacturing millions of sequences in parallel. The trade-off is that phosphoramidite synthesis relies on hazardous organic solvents and is typically run in specialized, centralized facilities—a model borrowed from semiconductor foundries, but applied to molecules.

Enzymatic DNA synthesis, in which DNA is assembled with the help of enzymes inside water, is widely viewed as a gentler, more biologically friendly route. It also more closely resembles the way cells build DNA naturally. Its limitation, until now, has been throughput: parallel enzymatic runs have lagged conventional manufacturing by roughly an order of magnitude. The Harvard chip’s 64-sequence figure is the team’s stated milestone against that backdrop.

The chip’s ability to address each site independently with its own ring electrodes is what enables the higher parallelism. Instead of one big chemical bath affecting every strand identically, each of the 64 positions gets its own miniature pH environment, switched on and off electrically.

An accidental route from brain recording to DNA synthesis

A notable detail in the reporting is the chip’s origin story. The underlying silicon electronics were originally designed by Jeffrey Abbott, a former PhD student in Ham’s lab, to record electrical activity from large populations of neurons. Those brain-interface chips relied on precisely controlled currents to permeabilize cell membranes and read intracellular signals.

Ham, quoted by ScienceDaily, said the group eventually asked whether the same current-control machinery could be redirected “from cells to molecules.” By swapping the neuron-facing electrodes for ring pairs that localize pH rather than membrane voltage, the team found the platform could also drive enzymatic DNA synthesis. The result is a case study in research tools migrating sideways into adjacent fields—in this case, electrophysiology tools repurposed for nucleic-acid chemistry.

Storing data, not just genes

To illustrate a long-term use case, the team encoded a 169-byte text across the 64 synthesized sequences. DNA data storage—the idea of archiving digital information in synthetic DNA—is attractive in theory because DNA is extraordinarily dense and stable over long timescales. In practice it remains uneconomic because writing DNA at the scale required (petabytes and beyond) would demand synthesis capabilities far beyond current facilities.

Co-first author Woo-Bin Jung, now at the Pohang University of Science and Technology (POSTECH), tied the data-storage demonstration to the synthesis approach directly, arguing that water-based enzymatic chemistry becomes increasingly attractive as production volumes rise, partly because cutting out organic solvents would shrink the environmental footprint of large-scale DNA manufacturing. Whether that promise translates outside the lab is one of the open questions the paper does not answer.

Why it matters

Synthetic DNA is a feedstock for routine laboratory work: PCR primers, CRISPR guide RNAs, diagnostic probes, gene-therapy constructs, mRNA vaccine templates, and libraries used in cancer and neuroscience research. Most of these uses are served today by centralized oligo vendors, which often accept hazardous-solvent waste as a cost of doing business. A distributed, water-based, electrically driven synthesizer—even a benchtop one—would change who can make DNA, where, and under what safety profile.

For diagnostics, the practical implication is the possibility of rapid, on-site synthesis of pathogen-specific sequences in outbreaks, without shipping samples or reagents through specialized facilities. For therapeutics, smaller and cheaper custom DNA could lower barriers for academic labs and clinical-trial groups designing gene-therapy vectors. For environmental policy, removing organic solvents from the synthesis chain would eliminate a regulated-waste stream that currently sits at the heart of DNA manufacturing.

For competitors to existing oligo-synthesis companies, the result is an early signal that the long-standing chemistry of phosphoramidite synthesis—first developed in the 1980s—may eventually be displaced by enzymatic alternatives. That transition would take years and would require independent replication and scaling, but the Harvard result adds to a body of work pointing in that direction.

Putting the numbers in perspective

The headline figure—64 parallel sequences, each up to 39 nucleotides—is impressive inside the enzymatic-synthesis niche but small compared with industrial phosphoramidite platforms, which ScienceDaily describes as routinely producing “millions” of sequences in parallel. The mismatch highlights where the field still stands: enzymatic synthesis is climbing from roughly a dozen parallel sequences toward dozens, while conventional chemistry already operates at a scale of millions.

The strand length is also modest. 39 nucleotides is enough for PCR primers and many CRISPR guide RNAs, but shorter than some therapeutic constructs. The team does not claim production-scale performance and frames the work as a milestone rather than a replacement technology. Comparisons drawn in any future coverage should be made against enzymatic benchmarks first, not against the entire synthetic-DNA industry.

Where the reporting converges and where it diverges

The ScienceDaily write-up is the only detailed source available; the underlying Nature Electronics paper is summarized through it. No independent technical commentary from outside Harvard is included in the source material, so claims about replication, yields, error rates, or cost-per-base have not been independently verified here. Readers looking for numbers on synthesis fidelity, reagent costs, or cycle time would need to consult the primary article or follow-up studies.

There is no contradiction within the available sources. What remains unconfirmed is the breadth of the result: whether the pH-localization strategy scales to hundreds or thousands of sites on the same chip, whether the 39-nucleotide ceiling reflects chemistry limits or engineering limits, and how the chip’s error rate compares with commercial oligo synthesis.

What to watch next

A few concrete signals would indicate whether this result is the start of a new synthesis platform or remains a laboratory demonstration. First, independent replication by a group outside Harvard—particularly one focused on chemistry yields rather than electronics—would test whether 64 parallel sites is a robust number or a difficult ceiling. Second, any movement toward a commercial spinout or licensing deal from Ham’s lab would suggest the technology is being taken beyond academic curiosity. Third, follow-up work pushing sequence length past 39 nucleotides, or pushing parallelism into the hundreds of sites, would mark real progress toward the production-scale benchmarks commercial users care about.

For now, the clearest takeaway from the available reporting is narrower and more durable: a silicon chip originally built to listen to neurons has been shown capable of directing 64 independent enzymatic DNA syntheses with electricity alone, using water instead of the hazardous solvents that have defined the field for forty years.