Over the past few years, many companies have moved lots of their computing jobs to a public cloud—specifically to Amazon Web Services—instead of running them on their own servers. It’s a popular technology and business model, one that Google Cloud Platform wants a large piece of. And with good reason: Amazon’s cloud computing unit alone is on track to surpass $10 billion in sales this year.
But this public cloud model has one big drawback: It doesn’t accommodate the many businesses that want to run some software in the cloud and keep some data in-house.
For these companies, Amazon (amzn) and Google (goog) don’t have a solution. That may soon change, as both companies are now quietly working on software that will run in customers’ server rooms, according to The Information (subscription required). That would, in theory, enable what techies call a “hybrid” cloud that mixes shared public cloud infrastructure with privately controlled data centers.
Data Sheet, Fortune’s daily technology newsletter.
Citing unnamed sources, the report acknowledged that the work might not come to market. Perhaps, but both of these companies eventually need to do something to in the arena of on-premises computing. The reason for this is that IT executives at many companies still want to keep key data and software to themselves. Even AWS chief executive Andy Jassy has acknowledged that not every company is going to shutter all own its data centers. And Google has brought former VMware (vmw) chief Diane Greene aboard to get a more corporate mindset for its cloud.
If Google (goog) and Amazon (amzn) execute well and quickly, they could hurt Microsoft (msft), the software giant already ensconced in corporate data centers that has long promoted its view of hybrid infrastructure. The company’s ambition is to get customers to run some software in the Microsoft Azure public cloud and some in their own data centers.
For more on Microsoft, watch [fortune-brightcove videoid=4929164340001]
However, a key piece of technology to enable that model, Azure Stack, originally promised by the end of this year, has been pushed back till mid 2017. Azure Stack is essentially a miniature version of the software Microsoft uses to power its Azure cloud data centers.
In July, aside from acknowledging the Azure Stack delay in a blog post, Microsoft said Azure Stack would only initially run on from Hewlett-Packard Enterprise (hpe), Lenovo (lnvgy), and Dell Technologies’ servers. That caused a ruckus among Microsoft partners and customers who want to deploy now—they think the delay could help competitors.
But Microsoft already has an advantage because many companies (of all sizes) already run Windows Server on their own hardware, giving it a big foothold in existing shops that AWS and Google do not have. Up until the last year or so, AWS executives seemed to pooh-pooh the need for on-site data centers, but that has started to change. The company seems to have realized that not every gigabyte of data nor every software application is going to move to shared infrastructure. It’s become clear that AWS, which is the leader in public cloud, will have to address this issue of dealing with, if not embrace, customers’ on-premises computing.
To be fair, Amazon does have its AWS Storage Gateway that sits in customer data centers, but it acts mainly as a funnel to move in-house data to AWS. Google had an enterprise search appliance that also worked within companies but discontinued it, replacing it with a software-only solution.
Fortune reached out to Microsoft, Amazon, and Google and will update this story as needed.
Note: This story was updated to include mention of AWS Storage Gateway and the Google Search Appliance.