There is no doubt that geopolitics is a major driver of TikTok’s fate in the United States. But if and when it does get sold, it’s still a multi-billion dollar transaction that has to make some business sense.

I have written quite a few posts on the various political and regulatory angles of TikTok already, so I won’t belabor those points here (though I do hope you give them a read for context of my general thinking). Instead, I’ll focus on how to think about what part of TikTok is worth, to whom, and why, in particular for Microsoft+Walmart and Oracle -- the two most likely groups of bidders to successfully buy TikTok.

Typically, an acquisition of any size needs to make sense along at least one, but hopefully more than one, of these dimensions:

  • Product synergy
  • Users and data
  • Core technology
  • Brand
  • Company and people

Let’s take a look at the relative value of each of these in the context of TikTok.

Product Synergy: Cloud and E-Commerce


The entire global cloud industry is still relatively young, growing quickly, and thus all the major vendors are in “land grabbing” mode, trying to gobble up market share first before worrying about profitability. But the landscape is also transitioning to “differentiation” mode, where cloud providers are looking for ways to “stand out”. While you can do a fair amount of differentiation with aggressive marketing, the most effective way to “stand out” is to demonstrate what types of workload your cloud can handle better than everyone else in the industry.

That’s where TikTok comes in. TikTok, as a product, has a unique set of workloads that requires:

  • storing and processing a ton of videos;
  • visual search with computer vision;
  • live streaming;
  • running AI algorithms continuously to serve different feeds and ads for different uses;
  • (increasingly) e-commerce transactions directly in the app.

Thus, it’s no surprise that TikTok uses both AWS and GCP as its two cloud providers in the US -- the same two providers that Snap uses. Being the “visual communications” company, Snap’s workload is the closest mirror to TikTok’s, even though it lags far behind in AI, live streaming, and e-commerce. (Snap spends handsomely on these two clouds, which I did a deep-dive on in “AWS and GCP Are the Real Winners in Snapchat’s Growth” a few months back.)

For either Microsoft or Oracle, acquiring TikTok and running the product on its cloud will help differentiate itself from AWS and GCP along these workload dimensions. This differentiation is not just a marketing practice, but a real stress test of engineering. It’s a stress test that both Microsoft and Oracle lack, because they have always been enterprise software companies with little or no experience running consumer applications at scale. Both firms are good at being “enterprise ready”, but not at being “Internet ready”. And the ability to handle workload stress at scale is what cloud customers pay for.

To proxy what kind of “stress” a cloud can handle, it’s helpful to analyze the company’s other core businesses and the workloads required to run them well. (As an example of this type of “stress test” analysis, see a two-part series I wrote a few months ago profiling all the major cloud platforms: AWS, Azure, GCP, Alibaba Cloud, IBM Cloud, Oracle Cloud, Tencent Cloud.)

Based on Gartner’s new global cloud market share analysis released in August, the top five players are: AWS, Azure, Alibaba Cloud, GCP, Tencent.


Everyone is still trying to catch up to AWS. As for the brand-name laggards who didn’t make the top-five cut, namely Oracle and IBM, they urgently need to catch up, both by continuing to invest in building data centers and product R&D, as well as “stress testing” its product to differentiate.

What’s why Larry Ellison aggressively pursued Zoom as a customer (and succeeded) earlier this year when Zoom was going through hypergrowth (see “Why Zoom Chose Oracle”). IBM should have been in the picture as well, but is likely busy absorbing its massive acquisition of Red Hat that just closed last July.

Having TikTok run on your cloud especially while it’s rapidly growing (100 million MAUs, 50 million DAUs, 800% MAU growth since January 2018 in the US), and run it well without downtime or performance degradation, is a legitimate accomplishment and extremely valuable for differentiation. After all, TikTok did reach 1 billion users faster than any social media app in history (even though it’s not really a social media app, more on that later).


The e-commerce synergy with TikTok is weaker than the cloud one in the short-term, but will become increasingly important in the long-term. Douyin, the Chinese version of TikTok, already has many e-commerce capabilities like buying items directly from a video or livestream feed. These capabilities are becoming table stakes in the Chinese e-commerce space, but are quite novel in the US and Europe.

Large retailers are all trying to transition into e-commerce ASAP to not continue losing market share to Amazon, and hopefully surpass it one day with new ways to shop. From that angle, the Walmart alliance with Microsoft to bid on TikTok is very forward looking. Microsoft and Walmart already have a tight relationship on cloud, with a 5-year partnership inked in 2018. Walmart is about to launch its Amazon Prime competitor, Walmart+, on September 15, coincidentally (or not) on the same day as the US government deadline to sell (or ban) TikTok. While Walmart+ is still a nascent product, merging TikTok’s e-commerce potential with it, or even just creating the perception of that happening, will do wonders to Walmart’s reputation as becoming an innovative tech company.

That does beg the question: why wouldn’t Amazon try to buy TikTok too? The truth is: Amazon has already been working with TikTok to allow users to buy directly on Amazon, advertising, and even content for Amazon Fire TV. Some of those partnerships will likely continue regardless of who ends up owning TikTok. Sadly, buying TikTok outright is a fool’s errand due to the simple reality that Trump doesn’t like Amazon. His disdain towards Jeff Bezos is well-documented. AWS is still fighting for a piece of the Defense Department’s JEDI cloud contract with a pending lawsuit alleging political interference from the Trump White House. There are already too many insurmountable hurdles.

As with all things related to TikTok, politics has a way of intruding.

Users and Data

Given TikTok’s staggering growth and the staggering amount of data being collected from its user base, acquiring those users and their data is obviously valuable; no insight here. And depending on what you read, TikTok’s data collection practice is either “just as bad” as the likes of Facebook and Google, or demonstrably worse.

What is worth thinking about is whether the acquirer has a good enough reputation to be entrusted with that data, whether it can derive value from the data, and keep growing the product to get more data.

Both Microsoft+Walmart and Oracle have decent reputations when it comes to data governance largely because their core business is either enterprise software or brick and mortar retail, where collecting user data is not at the center of their business model. This characteristic also makes them more trustworthy when it comes to national security concerns related to TikTok’s data. The irony of this reputation is that it’s built from their very ignorance in operating a product like TikTok. While it’s helping with the acquisition and various regulatory and public relations hoops the transaction will need to jump through, it also makes TikTok’s future business value and growth more murky.

Microsoft has had a bit more experience with consumer applications via acquisitions than Oracle, with LinkedIn and GitHub. However, LinkedIn is a first-generation social media product used primarily for enterprise sales or recruiting, not entertainment. GitHub’s status as a social network of the world’s developers is accidental at best. TikTok is not even a social media platform; it’s an algorithm-driven entertainment platform. I frankly don’t find Microsoft’s “experience” that relevant to its ability to run TikTok well.

The value of data is also tightly connected to the AI algorithms that make use of it, which is TikTok’s core technology. Is TikTok’s data more valuable than its technology?

Core Technology

I’m going to go out on a limb and opine that TikTok’s core AI technology is not worth as much as its users and data. This is one of those “strong opinions, weakly held”, but here’s why I believe that’s the case.

The concept of “data density” is what determines a lot of the value in a technology business. In the cloud industry, convincing your prospective users to migrate and store their data in your cloud is the most important step in growing long-term customer value. That’s why cloud storage is cheap (or comes with steep discounts), but compute resources and networking bandwidths are not. In e-commerce, the accumulative data on your shoppers’ preferences, behaviors, and habits is much more useful than the actual value of the items they buy. Same for an ad-funded social network. Same for a self-driving car company, whose cars need to self-drive millions of miles -- a “data density” building exercise -- before reaching the necessary level of safety and stability.

With “data density”, whichever combination of Deep Learning AI algorithms or frameworks (many of which are open sourced and commoditized) you apply to the data, you’ll get some powerful results.

Without “data density”, the algorithms will not be able to live up to their full potential, regardless of how “smart” or optimized they are.

This is not to say that TikTok’s algorithms are not valuable. Of course they are. Otherwise, the Chinese regulators would not be revising its own technology “entity list” to add “data analytics-based personalized information push service technology” as a new item. Because of this addition, whoever buys TikTok may not get to have its powerful algorithms. That’s obviously not ideal for the acquirer, and may drive the eventual sale price down. But the acquirer will have the users and data -- the more valuable piece -- to rebuild the recommendation engine if it really wants to. Given its advances in AI and strong partnerships with organizations like OpenAI (creator of GPT-3), Microsoft is most likely to succeed in this regard. Oracle...much less so.

You may reasonably ask: Facebook and YouTube both have a boatload of user data. Why didn’t they build something like TikTok to compete head on?

Facebook is certainly trying, first with Lasso (shut down) and now with Instagram Reels. Its inability to compete with TikTok and its predecessor does confound me. My best guess is Facebook’s DNA is social, where incentivizing new users to “friend” others is at the core of its product stickiness and growth strategy. TikTok is fundamentally different. You don’t have to “friend” anyone, “connect” with anyone, or even “follow” anyone to get some entertainment out of TikTok. It’s not that social. Thus, incubating a TikTok out of Facebook requires some fundamental learning and unlearning, which is hard for a company of that size to do.

As for YouTube, it does have the world’s largest catalogue of user-generated videos backed by the world’s most advanced AI company, Google. In short: YouTube was dealt a great hand to be TikTok before TikTok. But that outfit appears to be focused on showing as many ads as possible these days, especially now that it's a separate line item in Alphabet’s earnings. As for innovation, well...YouTube circa 2015 looks pretty much the same as YouTube today.

YouTube 2020 (black) vs YouTube 2015 (white)


A cool brand is extraordinarily valuable in the long run, but hard to quantify in the short run. TikTok is a cool brand -- a somewhat underlooked element of this whole transaction. It’s cool like Instagram, Snapchat, and Vine used to be cool.

Owning a cool brand can rejuvenate an old, not-so-cool company, which is a category that includes Microsoft, Walmart, and Oracle. (Although Microsoft has done an amazing job of rebuilding its coolness among developers by driving the popularity of the VSCode IDE, acquiring GitHub, and embracing open source -- a great story that I’ll write about on another day.)

All the possible acquirers could use a brand reboot. An effective image makeover could attract more talent, boost internal morale, and lead to more customers, so it’s worthwhile to pay a “cool premium”. How well it’ll work in the long run depends on how they will treat TikTok after the deal closes. I’m sure the boilerplate “TikTok will remain TikTok” line will be repeated on the public relations front, if only to not scare away TikTok’s core users, who skew young and have probably never used Windows, shopped much at a Supercenter, or even heard of Oracle RAC. But it’s a tough balance to maintain.

TikTok’s nemesis, Facebook, has had arguably the most experience in keeping a cool product untouched to maintain its coolness after acquisition. But even that could not last forever; a “From Facebook” splash was added to Instagram and Whatsapp last August and already appeared in Oculus.

Will “From Facebook” make Facebook itself more cool, or Instagram and Whatsapp less cool? It’s too early to tell. Generally though, I would not deem a product that has both teenagers and their grandmas cool, unless the grandma can do this.

Will we see a “Brought you by Microsoft”, “Powered by Oracle” or “Courtesy of Walmart” when loading TikTok one day? Your guess is as good as mine.

Company and People

The last dimension we will look at is “company and people”. Usually in an acquisition, even in a fire sale, there’s some value to acquiring the best technical talent and leadership who built the company, the so-called “acquihire”.

In TikTok’s case, this piece is probably the least valuable of all the five dimensions we’ve looked at. I mean no disrespect to the people who work at TikTok. By and large, ByteDance has a solid reputation for having cream of the crop engineering and product talent. Its US operation has been aggressively poaching top talent from the two meccas of tech talent, Google and Facebook. I presume the overall quality of talent is quite good.

But most of these personnel likely have not hit their strides to become productive. They came together during a year when TikTok was going through hypergrowth and when the proverbial geopolitical “shit” hit the fan. Assuming they haven’t quit, many of them not only needed to onboard properly in TikTok US, but also jell with their counterpart teams in China. Having run a much smaller cross-border team in the past, I know first-hand how difficult that “jelling” can be, even under the best circumstances.

All this chaos is happening while ByteDance is trying to increase its headcount from roughly 60,000 to 100,000 by the end of 2020, a hiring plan announced in March. To put this 40,000 person increase in perspective, Facebook’s entire headcount at the end of 2019 was about 45,000. ByteDance’s recruiting department is probably not going to hit its OKR this year.

As for leadership, TikTok’s shining white knight, Kevin Mayer, already quit. But he was always more of a “mercenary” than a “missionary” anyway. Whether it’s Microsoft+Walmart,  Oracle, or another acquirer, the “global role [he] signed up for” has become obsolete.

While this analysis is by no means comprehensive, I hope it gives you a good overview of what TikTok’s worth, to whom, and why, from a pure business perspective.

Geopolitical concerns are permeating the tech and business worlds these days, so much so that they get over-indexed in our thinking sometimes. This is not to say that there aren’t real national security concerns with TikTok, but there is a regulatory framework to deal with them, properly and separately.

At the end of the day, technology decisions still have to make technological sense. Business decisions still have to make business sense. Interconnectedness is everywhere, but sometimes, it doesn’t need to be.

If you like what you've read, please SUBSCRIBE to the Interconnected email list. New posts will be delivered to your inbox (twice per week). Follow and interact with me on: Twitter, LinkedIn.





  • 产品协同效应
  • 用户和数据
  • 核心技术
  • 品牌
  • 公司和员工





  • 存储和处理大量的视频;
  • 以计算机视觉为基础的视觉检索;
  • 直播;
  • 持续运行各种人工智能算法,给不同用户看不同的视频和广告;
  • (越来越多)直接在app里进行电商交易。


对于微软或甲骨文来说,收购TikTok再把它运行在自己的云平台上有助于和AWS和GCP相比时做差异化。这种差异化不仅仅是营销宣传部门的事,更是对基础工程团队的一种挑战和“压力测试”。这是微软和甲骨文都缺乏的一种压力测试,因为它们都一直是企业级软件公司,几乎没有或根本没有大规模运行 toC 应用程序的经验。两家公司都擅长“企业级”服务,但并不擅长“互联网级”的需求。而云客户买的就是能承受大规模工作负载压力的平台。





这就是为什么Larry Ellison在今年早些时候,当Zoom正在经历超速增长时,努力(也成功)的说服Zoom成为自己的云客户(参见《Zoom为什么选择了Oracle》)。IBM其实也应该参与其中,但可能正忙于吸收去年7月才close的对Red Hat的大规模收购。




大型零售商都在努力尽快过渡到电商领域,以避免市场份额继续输给亚马逊,也许还指望有一天以新的网购方式超越亚马逊。从这个角度来看,沃尔玛与微软结盟竞购TikTok还是有点前瞻性的。微软和沃尔玛在云计算领域已经有了紧密的合作关系,双方早在2018年就签署了长达5年的合作关系。沃尔玛即将于9月15日推出一款与Amazon Prime直接竞争的产品,Walmart+,这一天恰好是(或是故意的?)美国政府指令销售(或禁止)TikTok的最后期限的同一天。虽然Walmart+仍然是个新产品,但将TikTok的电商潜力与之相结合,或者起码创造这种未来的可能,都会对沃尔玛想打造成一家有创新的科技公司的声誉有好处。

这也引出了一个问题:为什么亚马逊不尝试收购TikTok?事实是,亚马逊已经与TikTok在合作,允许用户直接在亚马逊上购买东西,也与Amazon Fire TV做内容合作。不管谁最终拥有TikTok,一些合作关系可能都会继续下去。可惜,由于特朗普不喜欢亚马逊这一简单的现实,购买TikTok的这个想法无从起步。特朗普对Jeff Bezos的敌意已经众所周知。AWS仍在争取国防部JEDI云合同的一部分,诉讼未决,指控特朗普白宫对招标过程有政治干预。对亚马逊来说,已经有太多无法逾越的障碍了。






与甲骨文相比,微软在 toC 产品方面的收购及运维经验更丰富一些。毕竟买了LinkedIn和GitHub。然而,LinkedIn是款第一代社交媒体产品,主要用于企业销售或招聘,而不是娱乐。GitHub作为全球开发者的社交网络的地位充其量也只是偶然发生的而TikTok其实根本不是一个以社交驱动的产品。坦白地说,我并不认为微软的“经验”对它今后能不能好好做好TikTok有太大关系。









Facebook当然一直在尝试,先是Lasso(已经关了),现在是Instagram Reels。FB的各种产品无法与TikTok竞争,甚至与当年的Musical.ly竞争,还真是件让我困惑的事情。我最好的猜测是Facebook的DNA是社交,激励新用户与他人“交友”是其产品粘性和增长战略的核心。TikTok是根本不同的一种产品。你不必“加”任何人,“跟”任何人,就可以从TikTok里得到一些娱乐,与社交毫无关系。因此,从Facebook里孵化出一个TikTok需要一些根本的新学习和忘却,而对于这种规模的公司来说,还是很难做到的。


YouTube 2020 (black) vs YouTube 2015 (white)



拥有一个酷品牌可以让一个不那么酷的老公司焕发活力,这类公司里绝对包括微软、沃尔玛和甲骨文。(尽管微软通过推动VSCode IDE的流行,收购GitHub,并拥抱开源,在开发者群体里重建了它的形象,这个转型是个好故事,我会改天再继续分析。)

所有可能的买家都在某种程度上需要个酷品牌改造自己的形象。一个有效的形象改造可以吸引更多的人才、鼓舞内部士气和增加更多的客户,所以为“酷“多付点钱是值得的。但从长远来看,最终的效果如何还取决于交易结束后买家将如何对待TikTok。我相信 “TikTok会一直是TikTok” 这句“台词”将在公关层面不厌其烦的重复,哪怕只是为了不吓跑TikTok的核心用户:他们都偏年轻,可能从没用过Windows,也没在“沃尔玛超级中心” 买过多少东西,就更没听说过甲骨文RAC了。但这是一个很难维持的平衡。









至于领导才能,TikTok的“白衣骑士” Kevin Mayer 已经不干了。但其实他一直是个“雇佣兵”而不是“传教士”。无论是微软+沃尔玛、甲骨文,还是另一个买家,都不需要 “他签约的全球级别角色”。