We’re excited to announce that Modular has entered an agreement to be acquired by Qualcomm. The future of unified compute has never been stronger. Read the full announcement.
Join Qualcomm’s Investor Day today at 2:15 PM ET to learn more.
This announcement does not help the “Mojo is going to be a rug-pull” sentiment that many people had when Mojo was initially announced.
Today, the fact is Modular has been sold before v1.0 has been published.
Most programmers will not take on the risk of investing time and effort to build up the ecosystem of a programming language that is not in the control of its community.
The best thing Modular can do at this stage is to clearly commit to establishing an independent custodian for the language, like Rust and the Rust Foundation, for example.
@monte appreciate the input. We can confirm that we are still open sourcing Mojo this year as planned - we can’t wait to share the details at ModCon, August 18th in San Francisco. This will be live streamed. You can also read a bit more about Chris’s perspective on the acquisition and the plan to accelerate our open compute vision here.
We hear you, and thank you for all the questions. We can’t address them right now, but we will host a community meeting to be scheduled soon and focused on answering questions from the community. Stay tuned for more details.
What a silly thing to say. Maintaining and developing a language costs money and requires relationships with key partners, all of which are facilitated by such an acquistion IF the terms were set right (let’s see what the terms are).
Guido Van Russom with Python, Chris Lattner with Swift, React and Pytorch with Meta, and many others are examples of what a succesful OSS project backed by a Fortune 500 company can look like best case scenario.
There is chance of private interests capture but the Modular team have bought themselves a lot of trust and good will with me at least, for me to jump on the opportunity to call them sell outs.
Hello, Yordan.
To be clear - I am not accusing Modular of anything - “Mojo is going to be a rug-pull” is a sentiment that exists independently of my own views. This sentiment (fear) comes up in threads on public forums like Reddit and Hacker News. On Reddit, I have seen Chris Lattner respond to such sentiments.
I agree with you - let’s see what the terms are.
But it is important to understand that Qualcomm is not a partner - Qualcomm is set to own Modular.
Now, I trust Chris - we all trust Chris - otherwise we would not be part of this community.
Chris has made big promises about OpenSource, democratizing compute and building the language to V1 in closed-source, for the sake of building it to be the best it can be.
These are all great things, and I believe them to be true because of the passion that Chris and the team have.
The issue is -
Qualcomm is not beholden to the promises Chris made.
Qualcomm is beholden to their investors and their aggressive financial targets, as outlines in their ‘Investors Day’.
Therefor, realistically, we don’t know if all, some or none of the promises will be kept, in the short term or long term.
-
Python is owned by the Python Software Foundation (PSF), not a specific company, which is what I am suggesting for Mojo.
-
Both React and PyTorch originated from Meta, but they are now governed by independent, vendor-neutral non-profit foundations, which is what I am suggesting for Mojo.
-
Swift is owned by Apple, but has a very permissive Apache License 2.0.
I would venture that Swift sees less general adoption compared to Python/React due to not being governed by an independent foundation. (I know these technologies differ wildly and should not really be compared)
The companies you listed (Meta/Apple) are fundamentally software companies, which Qualcomm is not.
Qualcomm focuses on chip manufacturing (Qualcomm CDMA Technologies)
and patents (Qualcomm Technology Licensing).
I personally struggle to see:
- How a company that makes money on patents, will provide a permissive non-restrictive licence, and keep to that license in the long term if the underlying tech is proven to be of strategic value over competitors.
- How a chip manufacturer will refrain from private interests capture to benefit their own hardware above others with regard to execution of Mojo code.
Swift has a similar issue with Apple’s walled-garden, which adds to the poor adoption. - How a proprietary chip manufacturer aligns with democratizing compute.
These are valid concerns, and the type of concerns that will need to be addressed if Modular is to keep the trust that Chris has cultivated.
@conorbronsdon
I am honestly hoping for the best for Mojo
and I believe that is for Mojo to be governed by an independent, vendor-neutral non-profit foundation.
Good points!
I assume that the Python Software Foundation model is the way this went because I am not entirely sure what it would mean to own Modular if Mojo is open source and eventually Max (inference and training) as well.
If Modular can keep open sourcing their work while satisfying the needs of their new owners we all win and the only way it goes sour if we now push back the roadmap to Mojo phase 2 and 3 because Qualcom wants to focus only on edge inference.
Active support is not what Modular has been staffing for and their promise is hardware agnostic performance so I tend to think they will avoid becoming the tech support team for Qualcom.
I don’t think Qualcom acquired Modular to extract value from it. They already have a majority of the premium Android chip manufacturing, so I see it more as promotional, almost philanthropic acquisition, like Microsoft acquiring GitHub (hopefully with the opposite effects).
I also think Modular negotiated more for long term support and stability over a cash out and at their scale (4B) they are probably a minor investment for a 200B + company.
I guess we wait for more details in August 
Bruh! I think it doesn’t affect the fact of building the ecosystem rather it increases the leverage and intensity of how Mojo is taken.
Qualcomm is a great company which always lead’s in hardware manufacturing. Modular is simply broadening the horizon - This can’t affect Mojo’s contribution rather it accelerates so I see no problem in it.
Infact this will make Mojo’s porting to NPUs etc… like the Qualcomm TurboX easier and writing Kernels should be less obfuscated.
This is supercool this dude got no odds 