Microsoft Announces Q3 Earnings
microsoft.com
microsoft.com
Azure growth was inevitable based upon my (positive) experience over the last 3 months splitting an org into multiple tenants. Enterprise applications + app proxies are an amazing way to get your employees off legacy VPN crap and working with a simple org-wide SAML/MFA/SSO flow. And, this stuff actually works. It doesn't require invoking a bunch of arcane APIs to get the job done. I haven't seen anything short of a few Cloudflare offerings that gets anywhere close.
My most recent shit-test for Azure was building a FreePBX instance (i.e. non-windows) from install. It took me less than 1 hour to go from downloaded ISO to booting VM in Azure. The reason I made it from scratch is because I wasted an entire day trying to get AWS to export the old EC2 copy. I got something out but it definitely wasn't planning to boot up for me.
I am at a point where an "all-in" Microsoft stack is a super compelling idea. Today, when I go and look at Azure Active Directory, I have a level of confidence that simply does not exist anywhere else. Everything can be made to suffer our policies and procedures from top-down. Even other clouds. I've got AWS running as a sidecar to our AAD using SAML & SCIM to sync+auth identities. The only things keeping us from chopping AWS completely are Route53 (just domain registration) and S3 (because we are too lazy to learn something new at the moment). We currently have no direct exposure to Google products or services.
I would like to add one extra "FU" that I perceived throughout the Azure vs AWS experience: Microsoft has a browser plugin that will automatically run through the AWS control panel on your behalf in order to configure certain SAML/SCIM integrations. Clearly, at some point someone at Microsoft thought "this AWS UX is way too confusing for our customers" and actually cared about it enough to do something.
None of this even touches on Office, .NET, GitHub, Visual Studio, Xbox, etc. Each of these on their own would be a compelling reason to invest.
If I had a chance to make one argument to the board of Microsoft, I would plead with them to create a separate "hacker" edition of windows that is true price and ships with absolutely zero crapware. Some argue with me that Windows Server is effectively this, but brand image won't improve much with that narrative.
Adjusted for inflation, Windows 95 cost over $400 on launch day. I don't mind paying big bucks for an OS if it's done properly. This is precisely the same issue we run into with TVs. You'll find HN complaining about it approximately every 30 days or so. Price gets subsidized into a shitty experience every time. At least in the case of physical goods you have some realistic arguments for not differentiating the product line, but when it's a goddamn piece of software that already has a bunch of editions I don't see why not.
Microsoft strength was always their ability to understand and talk to their business customer. Microsoft always makes the effort to speak your language and bridge the gap between what you are looking for and what they offer.
Curious to see if they are going to end up eating AWS lunch or not.
My tiny little Azure Cloud nit on a rare-for-me encounters with Azure Cloud was the dev tools requiring two totally separate login types for seemingly no real reason other than a lack of updating a legacy system, and that made me forever think Azure Cloud was kind of dialing it in. It made working with them more complex than anything else I'd worked with.
One thing I think we agree a lot on is the value of good directory services:
> Today, when I go and look at Azure Active Directory, I have a level of confidence that simply does not exist anywhere else. Everything can be made to suffer our policies and procedures from top-down.
With my interest in DIY, self-hosting, I have long & increasingly felt like the weakest link is Directory Services. Tons of great great ways to run operate & maintain systems. But man, keeping track of employees (and ideally users too, albeit in a separate Directory) is hard! Roles are hard!
Red Hat's FreeIPA is good stuff (albeit on the somewhat off base/lower development 389ds Directory Server), a huge collection of tools to help setup infra. But it's been fairly inaccessible until semi-recently & it definitely requires significant growing in & expertise. And it's one of the only high-integration options out there. There's tons of great tools that complement each other well to build amazing directory service systems, but FreeIPA is one of the only already well integrated options.
And it's still only semi Cloud integrated. Trying to setup Kubernetes to use this all is still a huge second feat. I think, with the native cloud world, the directory service concerns are rapidly becoming the biggest gate, the worst hurdle, for those trying to set themselves up as capable computer operators. We seemingly both share appreciation for this task.
Microsoft has O365 (Dominant business tool), Github/Vscode (Dominant coding tools, also perfect datasource for AI coding), Gaming (Windows is still overwhelmingly dominant in gaming, and also Xbox exists), AI (They own 50% of OpenAI), and finally Azure.
Alphabet has the consumer google suite (Youtube/gmail/google) that just prints out extremely high margin cash. They also have android/chrome as defensive moats for the core suite.
Hence Amazon only beats out Meta in stock valuation
If a significant percentage of AWS customers are using AWS EC2 instances just as IAAS for Microsoft products then the AWS market share lead is primarily due to the inertia of Amazon's MBAs having figured out how to financially capitalize x86 machines for 0% interest rates before Microsoft's MBAs did (or more likely Microsoft's historic margins were so high that they didn't want to take the margin hit).
Never could find a number for it though...
[1] https://aws.amazon.com/windows/products/ [2] https://www.youtube.com/watch?v=HfGcrV9_3kQ AWS re:Invent 2020: Win the market opportunity for Microsoft workloads on AWS
This is exactly us. We picked AWS because it was the only thing that fit back in ~2015. Today, the landscape is dramatically different. Given our chance to re-evaluate, we decided the new org is going into Azure. I think the #1 reason we picked AWS was the integration between EC2 and Route53 (i.e. VMs and DNS management). Our workload previously consisted of 2 linux machines, and when we moved to Azure that dropped to 1. If we went even further into the Microsoft rabbit hole, that final linux box could be dealt with.
We made an executive decision/policy at the beginning to never use an AWS-only special to build our product or business. At all points we try to maintain a stack that should be compatible on any hosting provider. We fail with things like S3 usage, but there are alternatives we could realistically switch to if it became a serious pain point. Converting an AWS lambda stack into an Azure ??? is less confidence-inspiring for me.
Azure & Cloud probably are decent comparisons to each other, though.
We have also found that we are spending almost 25% more in Azure for similar compute capabilities as we've had to launch more servers to meet expected baseline performance (or pricier systems).
The lack of proper AZ's in certain locations is also a huge bummer, on top of that, we've found that in certain regions you will have certain instance types that are available in one AZ but not the other. So you end up having to find somewhat equivalents and rewrite infrastructure as code to deal with that mess.
Last but not least is the fact that GCP and Azure both don't have great support for IPv6. AWS may have been late to the game, but it's easy to deploy, and it just works. Whereas in GCP you have to be in certain regions, and in Azure I still am not sure what they were smoking with giving such a tiny public IPv6 prefix: https://learn.microsoft.com/en-us/azure/virtual-network/ip-s...
Azure's terraform provider is also not the greatest, unfortunately.
Been using AWS and Azure for a long time and only recently started using GCP.
I was surprised how fast GCP services and resources startup.
On AWS/Azure it takes more than a few for VM to spin up but on GCP it feels like a few seconds.
I wonder how much Microsoft is benefiting in cloud sales from the 'halo effect' of Bing/OpenAI/recent AI product announcements.
Google by contrast is looking at a sinking market share because they've missed the boat so hard and managed to continuously embarrass themselves by rushing out half-baked presentations and products.
However, I'm guessing those 'legacy' products are also benefiting from some positive brand association based on the last 3-6 months or so. Basically, even if they haven't yet realized meaningful revenue directly from Bing Chat/LLMs/AI-enabled products, they may already be capturing some meaningful indirect benefits through their other products.
Google overtook Microsoft in sales five or so years ago. So for example in fiscal 2019 Microsoft had $42b in operating income, and Google had $35b in operating income in fiscal 2019 - that's despite Google doing $36b more in sales that year (not quite 30% more).
For the last four quarters (excluding today's results) Microsoft has a ~41% operating income margin. Google by contrast has a 26% operating income margin (how much operating income are they generating from revenue).
On Google's $282 billion in sales the last four quarters, their cost of revenue is equal to ~45% of revenue. By contrast, Microsoft's figure is ~32%.
Microsoft has the far superior business imo. To say nothing of the risk that Google's core search business is about to get murdered by an AI inflection point that Alphabet screwed up (got lazy and extractive focused) and severely botched the transition on. For years Google has been too oriented toward financializing what they already had and they've been unwilling to try to kill off their own product by creating the future (the next search paradigm). It's pretty classic, the lazy monopoly gets capsized while resting on its laurels.
Alphabet has mediocre leadership for this time and place. Sundar Pichai is to Alphabet as Ballmer was to Microsoft. Perfectly fine managers (both managed considerable expansions, courtesy of inherited momentum), but that's all they can do. They're the wrong types to have in charge during a critical tech inflection that threatens the overall business. Nadella was the right person to have in charge of Microsoft to make the needed changes; and despite being behind the curve due to Ballmer, they were still able to recharge Microsoft and right the ship. The sooner Alphabet gets rid of Pichai the sooner they can begin attempting to avoid the beating that is bearing down on their core business rapidly. They're going to have to be willing to cannibalize the Google business as needed and Pichai can't pull that off (and the founders are probably not remotely hungry enough to want to try to do it, to put in the sweat necessary, based on their last years in management there and how they left). Amusingly Alphabet is facing this threat similarly as Microsoft did, under the very bright lights of anti-trust. It's probably not far-fetched to say that if Ballmer had remained in charge for another 5-7 years, it might have been too much, too far gone to recover from (they would have missed the ramp to the next era by too much); I believe the same is true about Pichai and the present leadership at Alphabet, they can't do what needs to be done, and if the board/large vote holders wait five or more years to make the needed changes, it'll probably be too late to fully recover from.
But I am quite optimistic of the application of LLMs in other products like Office.
It's possible that Google's search quality, which has already been trending down, completely nosedives when the internet gets filled to the brim with hallucinations.
I asked chatgpt and in a single response it gave me the metric I was looking for and citations for me to check. It was a much better experience.
I’ve also found I don’t need to google for programming questions as much and have fewer visits to stackoverflow.
I don’t know if it’s going to kill search entirely, but it’s definitely a bad thing for search.
Did you check the citations—that they existed, and said what ChatGPT claimed? It knows how to properly report metrics and format citations, but there's no guarantee that those metrics or citations are real, just because they're properly formatted.
Microsoft is just using a different year definition.
Some companies name their quarters based on the calendar year that the fiscal year starts in, some do it based on when fiscal year ends.
It probably matters to people in finance, but I'm not sure it matters to anyone else.
> A fiscal year is a one-year period that companies and governments use for financial reporting and budgeting. A fiscal year is most commonly used for accounting purposes to prepare financial statements. Although a fiscal year can start on Jan. 1 and end on Dec. 31, not all fiscal years correspond with the calendar year. [1] [2]
I'm not totally sure if this actually makes sense and is net positive for the business. But this is the reasoning I've heard from sales people for wanting to offset quarter schedules.
A lot of sales and cash flow events are seasonal in nature. Sales for a company may slow down during the summer and may pick up in the winter.
For example, a school year tends to run Sep - May, so any business that services the academic industry would probably also want to set their fiscal calendar to July thru June, so that sales, cash flow and income aligns with the actual sales cycle, rather than split their income statement across two academic years if they went with Jan thru December.
Just one of many examples. Another company might make most of their sales towards the end of the year but since customers buy on credit they might not receive the cash flow until Jan, Feb, or March (under a net 90 invoicing scheme). So they may choose to then have Jan-Mar be their Q4 so they can collect on that revenue to influence their EOY balance sheet, and to give them more of a buffer to cover EOY taxes.
"Because we're committed to providing you with high-quality chat responses, which are computationally expensive, we require users to sign in or create a free account to continue chatting with Docalysis. Thank you for understanding!"
This is not fair-play and someone should at least raise it up to level of awareness.