跟讀練習: Microsoft Is (Finally) Getting What They Deserve... - 通過影片學習英語口語
選擇練習模式
每種模式都會開啟全屏練習室。
86 句
如果句子過短或過長，請點選 Edit 進行調整。
1
On January 28th, 2026, Microsoft reported earnings that beat expectations on both revenue and profit, and the stock still dropped 10% in a single day, wiping out over $350 billion in value.
2
That reaction tells you something important.
3
Investors weren't reacting to what Microsoft reported, they were reacting to what Microsoft revealed.
4
Because buried in those results was a much bigger story.
5
Microsoft is pouring tens of billions into AI infrastructure at a time when growth is already starting to slow, all while relying on a strategy that's deeply tied to a partner whose own future is becoming increasingly uncertain.
6
Let's go back to January 2023, when Microsoft announced a massive new investment in OpenAI worth $13 billion.
7
Everyone called it the deal of the decade.
8
What they didn't tell you is that a significant portion of that investment wasn't even cash, it was Azure credits.
9
Microsoft was handing OpenAI vouchers redeemable at Microsoft's own data centers, booking the arrangement as investment, then also collecting the revenue when OpenAI used those credits.
10
It's the financial equivalent of a landlord paying a tenant's rent to himself, calling it an investment and counting both sides of the transaction.
11
Microsoft structured the deal so that they would collect 75% of OpenAI's profits until they had fully recouped their $13 billion investment.
12
Analysts at Morgan Stanley looked at OpenAI's growth trajectory and projected Microsoft would hit full recoupment sometime in 2026.
13
The problem with it was that it requires OpenAI to actually generate profits.
14
OpenAI lost roughly $11.5 billion in a single quarter in 2025.
15
You can't collect 75% of profits that don't exist.
16
And because Microsoft accounts for its OpenAI state using what's called the equity method, those losses hit Microsoft's own books.
17
They disclosed a $4 billion pre-tax hit in a single quarter.
18
So that recoupment timeline is now dead.
19
Instead of collecting profits, Microsoft is absorbing losses every three months.
20
Then, on January 21st, 2025, 5.
21
Altman stands at the White House with Oracle's Larry Ellison and SoftBank's Masayoshi Sun to announce the Stargate project.
22
That same day, Microsoft disclosed it had lost its status as OpenAI's exclusive cloud provider.
23
Then the deals started stacking up.
24
OpenAI signed a $300 billion deal with Oracle, a $22 billion deal with CoreWeave, and then the kill shot, a $38 billion seven-year deal with Amazon Web Services.
25
By the time the dust settled, Microsoft still held about 27% equity in OpenAI,
26
which sounds significant until you realize OpenAI had signed infrastructure deals across seven different vendors, and Microsoft was just one of them, capturing roughly 22% of where OpenAI's money was actually going.
27
They went from being the exclusive partner to being one option on a list.
28
The relationship that was supposed to define the AI era had been quietly shopped around to everyone including Microsoft's biggest competitor.
29
Microsoft has 440 million commercial 365 subscribers.
30
These are people who already live inside Word, Excel, Teams and Outlook every single day.
31
They built an AI assistant directly into all of those products and charged $30 per user per month.
32
On top of existing subscriptions to unlock it, only 2% of their entire subscriber base paid for it.
33
A Gartner survey found that of organisations that had finished co-pilot trials, only 5% were moving to larger deployment.
34
But the But the adoption numbers are almost a secondary problem, because the product itself has been generating headlines for all the wrong reasons.
35
Let's talk about Recall.
36
Microsoft announced it at their developer conference in 2024 as the flagship feature of a brand new PC category.
37
Recall would take a screenshot of your screen every few seconds, run it through AI and let you search everything you'd ever seen or done on your computer.
38
They called it a perfect memory.
39
researchers called it a keylogger built into the operating system.
40
And because Recall stored all of those screenshots in a local database that was unencrypted by default, a developer built a tool that could extract every single screenshot from that database within days of the announcement.
41
Microsoft pulled the feature before it shipped, delayed it an entire year, added encryption, and still had to address security vulnerabilities after it finally launched in April 2025.
42
by early 2026 fewer than 10% of Windows 11 PCs could even run it.
43
Then there was EchoLeak, a zero-click prompt injection exploit in Copilot discovered in June 2025.
44
A single email could silently pull data from your SharePoint, OneDrive and Teams through Copilot's AI context, with zero interaction from the user.
45
Microsoft patched it, but security researchers noted the underlying attack technique had been publicly documented for over a year before the exploit was discovered in production.
46
Microsoft spent years pushing Copilot into places users never asked for it.
47
A button in Notepad, AI suggestions in Photos, prompts in File Explorer, a dedicated hardware key on new keyboards.
48
Each update seemed to add a new place where Copilot would appear and offer to help with something you were already doing fine without it.
49
Microsoft's Windows lead eventually acknowledged publicly that the operating system had gone off track with aggressive AI expansion while basic stability issues went unfixed.
50
In March 2026, TechCrunch reported Microsoft was rolling back Copilot integrations across Windows.
51
The executive who announced it said the goal was to integrate AI where it's most meaningful, which is a careful way of admitting they had been integrating it where it wasn't.
52
The trust problem runs deeper than bad features.
53
A net promoter score tracking Copilot's accuracy went from negative 3.5 to negative 24 in the span of two months in late 2025.
54
Among users who stopped using Copilot after trying it, 44% said distrust of the answers was the primary reason.
55
What makes this particularly uncomfortable is that industry observers have been saying for years that the AI push inside Microsoft has less to do with user demand and more to do with justifying the massive capital investments already made.
56
Let's take a step back and look at where Microsoft actually stands in the AI landscape.
57
Search was supposed to be the first big win.
58
Bing chat launched in February the 2023 and Nadella told The Verge he wanted to make Google dance.
59
Three years later Google has 90% of global search share.
60
Bing has 4.3%.
61
Microsoft spent a decade of investment and $13 billion to move that number by about one percentage point.
62
In mobile AI, Google's Gemini crossed 750 million monthly active users.
63
Apple chose Gemini as the foundation for a rebuilt Siri across 2 billion devices in a deal worth about a billion dollars a year.
64
Microsoft is not in this picture at all.
65
They have no mobile operating system, no consumer social network, no significant browser share.
66
Their only consumer service is Windows and we just covered how that's going.
67
In developer tools, GitHub Copilot was supposed to be the one clear win, but a competitor called Cursor hit 2 billion in annual revenue by early 2026.
68
And according to one industry analysis, Anthropic captures 54% of the coding AI market and OpenAI about 21%.
69
GitHub Copilot is somewhere in the remainder.
70
Microsoft paid $650 million in 2024 to bring in Mustafa Suleiman, who co-founded DeepMind and previously served as CEO of Inflection AI to lead their in-house AI efforts.
71
Two years later, his team has produced models that, by his own admission, will need another year or two to reach frontier capability.
72
And while all of this was happening, three of the key people who built Microsoft's AI ambitions quietly walked out the door.
73
The architect of Bing Chat, who went to Shopify, the president of Azure AI, who went to Anthropic, and the lead researcher behind the Fi models who went to open AI.
74
By March 2026, Microsoft narrowed Suleiman's role and brought in a new executive to take over Copilot product, which is a fairly direct signal that the dual mandate wasn't working.
75
A leaked internal comment from a senior Microsoft executive described most of their Copilot products as gimmicky.
76
Another insider said, there is a delusion on our marketing side where literally everything has been renamed to have Co-Pilot in it.
77
By that point, Microsoft had somewhere around 80 distinct products, with the word Co-Pilot in the name.
78
The BBB's national advertising division ruled in June 2025 that Microsoft's Co-Pilot branding was misleading, citing unsupported productivity claims.
79
Microsoft put in $13 billion and lost board influence, lost cloud exclusivity, lost the the consumer, lost the developer, and are now absorbing billions in losses every quarter from the company they funded.
80
They built a product that 98% of their own existing customers declined to pay for.
81
They pushed AI into every corner of their operating system until their own executives admitted it had gone off track.
82
And when the market finally looked at all of that together on January, it wiped out $357 billion in a single session.
83
Microsoft spent more than almost anyone, moved faster than most, had advantages nobody else had, and they are still watching from the outside as the platforms that will define this decade take shape without them at the centre.
84
The question was never whether Microsoft could spend on AI.
85
It was whether spending would be enough.
86
And right now, the answer looks like no.
FOR YOU
推薦影片
C12
128 句
The AI Take Over Has Completely Backfired (Here's the Proof)
Business EnglishUniversity StudentsWorking Adults
C27
194 句
Github are you joking?
Business EnglishUniversity StudentsWorking Adults
B21
267 句
ALL OF BUSINESS Explained in 20 Minutes
Business EnglishUniversity StudentsWorking Adults
C211
101 句
Why SpaceX Will Become the World's Most Valuable Company
Business EnglishUniversity StudentsWorking Adults
B22
275 句
Mac vs Windows - It's not close in 2026!
British AccentBusiness EnglishUniversity StudentsWorking Adults
C23
106 句
Satya Nadella demos an app he built | Microsoft AI Tour Bengaluru
Business EnglishUniversity StudentsWorking Adults
C14
123 句
Margin Call - "Sell it all. Today." 👆🤘👆
Business EnglishUniversity StudentsWorking Adults
B27
179 句
OpenAI: A Bubble Bigger Than Dotcom
Business EnglishDaily ConversationUniversity Students
C13
83 句
Dan Ives: Apple had to raise prices and this was the right time
Business EnglishUniversity StudentsWorking Adults
C13
176 句
Is Amazon Too Big?
Business EnglishDaily ConversationUniversity Students
C12
163 句
A Tale of Two Markets
Business EnglishUniversity StudentsWorking Adults
C27
306 句
Beyond Code Coverage: Functionality Testing with Playwright — Marlene Mhangami, Microsoft
Business EnglishUniversity StudentsWorking Adults
為什麼要通過這個影片練習口語？
在學習英語的過程中，口語表達能力至關重要。這個影片展現了微軟在人工智慧領域的戰略變化和市場反應，提供了豐富的口語練習內容。通過模仿影片中的對話，學生不僅能夠提高自己的發音和語調，還能深刻理解相關的科技和商業用語。
使用“shadow speak”技術，學習者可以在觀看影片的同時，對照字幕進行練習，從而增強語感與記憶。經常通過看YouTube學英語，能夠幫助學習者更加自然流暢地掌握口語，這是一種有效的英語口語練習方式。
語法與表達方式分析
- 過去時態的使用: 影片中使用了大量的過去式，如“微軟報告了...”和“分析師預測...”。掌握過去時態的使用，可以幫助表達與過去相關的事件和事實。
- 條件句: “如果OpenAI沒有盈利，微軟就無法收回投資。”這種句式在商業環境中十分常見，學習如何構造條件句將有助於表達可能性或假設情境。
- 被動語態: “功能被取消”與“資料顯示...”。被動語態是學術和商務語境中常用的結構，可以有效地強調動作的承受者，而不是執行者。
- 並列句和複合句: 影片中頻繁出現的複合句讓對話更為豐富，理解並熟練運用這些句型有助於提升口語的複雜度。
常見的發音陷阱
- “Microsoft”中的重音: 確保在發音時重音落在第二個音節上，這對於非母語者來說可能會有些挑戰。
- 科技詞彙: 如“AI”（人工智慧）和“cloud provider”（雲服務商）等詞彙，學習如何正確地發音，並在對話中流利使用。
- 連讀現象: 在快速的語速中，比如“upgraded in”（升級為），連讀可能使非母語者感到困惑，練習時需特別留意。
通過這些練習，學習者可以在日常對話中更加自信地使用相關表達，提升自己的英語口語能力。利用“看YouTube學英語”，可以在家自行學習，省時省力，也開闊了學習的視野。如果你尋找的是一個shadowing site來進行英語口語練習，不妨加入這個影片中練習你的英語口語。
影片中的語法
說話人最常用的結構，並附上影片中的原話：
| 結構 | 影片中的用法 |
|---|---|
| 現在完成進行時 have/has been + -ing — 以前開始、現在仍在繼續的動作 | has been generating · have been saying |
| 定語從句 who / which + 從句 — 補充說明人或事物 | people who already · meaningful, which is · people who built |
| 被動語態 be + 過去分詞 — 強調發生了什麼，而不是誰做的 | was unencrypted · was discovered · has been renamed |
| 現在完成時 have/has + 過去分詞 — 過去發生但與現在仍有關聯的事 | has produced · has been renamed |
什麼是跟讀法？
跟讀法 (Shadowing) 是一種有科學依據的語言學習技巧，最初開發用於專業口譯員的培訓，並由多語言者Alexander Arguelles博士普及。這個方法簡單而強大：您在聽英語母語原聲的同時立即大聲重複——就像是一個延遲1-2秒緊跟說話者的影子。與被動聽力或語法練習不同，跟讀法強迫您的大腦和口腔肌肉同時處理並模模擬實的講話模式。研究表明它能顯著提高發音準確性，語調，節奏，連讀，聽力理解和口語流利度——使其成為雅思口語備考和真實英語交流最有效的方法之一。