Monday, July 01, 2019

Will Regulators Scotch PacBio Acquisition?

When Illumina announced its proposed acquisition of Pacific Biosciences last fall, an immediate question arose as to whether the deal would pass antitrust review.  Illumina must have been optimistic, as they originally expected the deal to close in "mid-2019".  That timetable has now been rolled back to late 2019 in the face of the UK agency with jurisdiction over mergers deciding to go for a second round of review and no word yet from the U.S.'s Federal Trade Commission.

I've had a lot of inquiries about this and had a few discussions, which has served as a little bit of schooling on this.  A very good thing.  For to use the expression my son has come to loathe, I don't remember any antitrust theory from all the time I spent in law school.  Just as I've forgotten what I learned about siege tactics during all my time in military academy. Don't even ask about Divinity School.  So remember in all of this you have a rank amateur in charge, so feel free to contest and recheck as you see fit.

A core concept for antitrust is a "theory of harm" -- how might Illumina act post-merger in ways it can't currently behave that would measurably harm consumers.  From discussions and reading a treatment written by an actual authority, I can boil these down into two major (and not mutually exclusive) classes of theory. The first class is concerned primarily with how an independent PacBio (or PacBio in someone else's hands) would potentially compete with Illumina; the second is focused on how Illumina might use the combined entity to squash other players.

Theory Class I: Removing a Future Threat

Smothering a nascent competitor is one class.  The extreme form can be seen in a question I've been asked by multiple people: "can you imagine Illumina shutting down PacBio technology after the acquisition?".   That version of the theory makes little sense to me; I think Illumina sees PacBio as complementary technology and there was apparently for quite a long time a "buy PacBio" contingent within Illumina.

But another version of this class of theory is the idea that PacBio was going to be a serious head-to-head competitor with Illumina in the near future and would therefore put pressure on Illumina to be less aggressive in pricing.  Conversely, post-merger Illumina can act like a cartel and not pit the two technologies against each other.  

Clearly there are areas where long reads from PacBio or ONT can currently compete with short reads from Illumina.  It's long been true that good microbial assemblies require long reads, though typically you would need high coverage Illumina data for error correction along with that.  Now with HiFi reads that might not be true, though certainly with R9.x nanopore reads the Illumina backstop remains critical.  With the 8M chip on the Sequel 2, the cost of generating either long insert reads or HiFi on PacBio has dropped dramatically, perhaps enabling a human genome for well under $10K (I don't remember seeing a cost claim).  But that still leaves a significant cost gap between resequencing on NovaSeq and de novo sequencing on Sequel 2.  I think there's going to be a lot of de novo long read sequencing of genomes, with perhaps crop plant germplasm for breeding outnumbering human genoems, but is it really going to put a serious dent in Illumina's business?  I doubt it.

Of course, it gets complicated.  There's going to be growth in other areas of Illumina's short read business.  In particular, it is my repeated opinion that counting applications will see spectacular growth and are much better suited for the current short read platforms than any current long read platform.  Counting emphasizes number of fragments read, not total basepairs.  PacBio is obviously a non-starter here; 8 million is rounding error on 8-10 billion fragments readable on a NovaSeq S4.  Trying to figure out what a PromethION can do here really can't be done on paper, as trying to go for many fragments means short fragments which historically means rapid flowcell degradation.  If anyone has empirical data in the field for PromethION fragment counts, I'm all ears.

Theory Class II: Throwing Elbows

Under the second class of theories are ways in which a post-acquisition Illumina could use their position to muscle aside other competitors in the sequencing market.  In other words, how might Illumina use their near monopoly in short reads to achieve a similar standing in long reads. Note that under some of the behaviors I'll outline, consumers benefit in the short term but are ultimately penalized: it's fun to have someone else foot your bills until they have you trapped and then extract from you with interest.

Bundling Illumina and PacBio products together is a rational business strategy, but at what point does it become anti-competitive?   Presumably lumping ones long and short read orders together to qualify for a larger discount is kosher, whereas Illumina offering to throw a free Sequel 2 in with every 5 NovaSeqs might not be.  Actually, I would expect competitors to insist that any sort of bundling is an issue, as it does leverage Illumina's enormous existing business to make PacBio extra competitive from a price standpoint.

As another example, the BaseSpace cloud environment might in the future offer seamless moving between PacBio and Illumina data, such as building a reference transcriptome with Iso-Seq and then quantifying a large number of samples with short read RNA-Seq.  Or build a reference genome with HiFi and then type the SNPs over your breeding population.  Even just integrating control and monitoring of all sequencers within BaseSpace could make buying only Illumina attractive for shops with limited resources for building their own infrastructure of this type.

An even more extreme behavior would be dumping, in which Illumina uses their profits from short reads to heavily subsidize PacBio purchases.  Forget buying the NovaSeqs, we'll just give you a Sequel 2!  Something like that, or giving away reagents to gain market share.

A significant concern stemming from any such behavior would be reduction in innovation activity in the space.  If Illumina could pick off each competitor in turn, which in military parlance is called a "defeat in detail", then they could charge monopoly prices once the dust settled.  Or they would just scare anybody from even financing a challenger.  Not the same thing, but at least one failed long read company blamed their inability to gain financing on some of the early buzz around Oxford Nanopore.

Possible Regulatory Remedies

How might regulators respond?  There are a number of different avenues that regulators can pursue, obviously starting with doing nothing for now.  They can also force the company to agree to meet certain conditions or refrain from certain decisions, often via a document called a consent decree.  Regulators can force the divestiture of components of the businesses.  In retail mergers, it isn't uncommon to see individual outlets in certain geographical areas to be divested as part of the regulatory component of a merger.  Or there's the nuclear option: blocking the acquisition.

Note that any sort of regulatory remedy can have unexpected consequences.  Consent decrees between the U.S. government and AT&T prevented the latter from expanding into computers.  As a result, software generated by AT&T Bell Labs could not be sold by the parent company and thus wasn't aggressively commercialized.  This is a significant reason why UNIX was initially an academic operating system: AT&T could give it away but not sell it. 

Might the U.S. or U.K. agencies insist on putting in place a consent decree enjoining Illumina from certain behaviors?  For example, perhaps BaseSpace would be required to support alternate platforms, which right now means Oxford Nanopore, at some level.  Defining what would constitute appropriate support could be a sticky point: is just ensuring nanopore data can get in sufficient or must nanopore-only tools be enabled?  Straight-up bundling and dumping could be prohibited, with some provision for accountants to review the financial terms of any sales to ensure this isn't happening.

Given that PacBio has only one product line and Illumina isn't going to give up short reads, the opportunities for a forced divestiture seem very limited.  The only I can think of would be a requirement to spin out BaseSpace as a fully independent company or sell it off. 

A grim reality is that PacBio is not a healthy company on the financial side.  If we take their most recent quarterly statement and divide cash ($38M) plus investments ($45M) by the quarterly loss ($30M), they have a bit less than three quarters of  a year  of runway.  If the merger is nixed by regulators, PacBio will need to scramble quickly to find another partner and would probably need some sort of financing just to survive another regulatory review period (whether the partner could provide that was yet another concept I failed to learn during the sum of my time in law school and business school).  

So it's a bit of white knuckle time for PacBio.  Further delay in approving a merger could be a problem.  There would be little time to negotiate conditions if ones were proposed by regulators.  The uncertainty around the company could also extend to groups considering buying into the platform. PacBio has developed a useful technology and I can't imagine it just disappearing in a financial cloud of dust, but there could be a disruption in the company that worsens any current brain drain.  Drawing out the process has no good effects for PacBio or for Illumina.



Anonymous said...

Hi Keith, always good to see your take on NGS issues.

One point above caught my attention; you wrote "at least one failed long read company blamed their inability to gain financing on some of the early buzz around Oxford Nanopore"

Any details on what company this was or how their tech worked?

Liang Zong said...

Thanks for the inspiring thoughts. If you were someone who in charge of ONT or other sequencer vendors, how would you take advantage of this delay? Thank you.

Anonymous said...

There’s about a 20% chance the merger is cancelled. 45% chance it goes through no restrictions. A good chance restrictions are applied and it goes through, a strong one is divestment of IP, where that IP impinges on existing competition. Another is PacBio is forced to be run as a separate entity, allowing any uncompetitive commercial practices to be transparent and so open to further action.

Anonymous said...

With regard to "at least one failed long read company blamed their inability to gain financing on some of the early buzz around Oxford Nanopore", I'm pretty sure it was Halcyon Molecular. When seeing Oxford Nanopore's presentation, one of the cofounder brothers (the Andregg's) immediately walked out and shut down the company. They were backed with PayPal mafia money via Peter Theil, Elon Musk, Luke Nosek etc. Interesting platform that may have worked but would have had major commercialization issues.
They were a weird group...many people in the Bay Area sequencing space have run into them. They still owe me most other people they've met.

Anonymous said...

In a rational world Clive Brown's big mouth should be enough to "prove" that PACB is not a threat to ONT. Anyone remember the asinine "I can smell fear" tweet, among many others? But now it seems that the tables have turned. We can smell your fear, Clive....

Anonymous said...

"In a rational world Clive Brown's big mouth should be enough.."

hang on. if this post is from a PacB person it shows the exact opposite, its your fear on display, scuttling off to hide behind ILMNs skirts, then gloating about it triumphantly ... implying you will automatically crush any competition ....

Anonymous said...

I guess we don’t live in a rational world.